Download UNIVERSITY OF CALGARY Direct, indirect and predator

UNIVERSITY OF CALGARY
Direct, indirect and predator-mediated effects of humans on a terrestrial food web:
implications for conservation
by
Tyler Bryon Muhly
A THESIS
SUBMITTED TO THE FACULTY OF GRADUATE STUDIES
IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE
DEGREE OF DOCTOR OF PHILOSOPHY
FACULTY OF ENVIRONMENTAL DESIGN
CALGARY, ALBERTA
OCTOBER, 2010
© Tyler Bryon Muhly 2010
ABSTRACT
Humans influence the density and distribution of other species through direct and
indirect effects. Effects are indirect when a species influences another species via an
intermediate species, whereas direct effects have no mediator. Humans have strong
negative direct effects on wolves (Canis lupus) in areas where wolves kill livestock. Such
effects could indirectly influence other species in food webs, particularly wolf prey (i.e.,
herbivores such as elk; Cervus elaphus) and vegetation species eaten by herbivores. In
addition, humans simultaneously directly influence several species in food webs,
triggering numerous indirect effects. For example, humans have direct effects on forage
(e.g., through agriculture) that could indirectly influence herbivores (e.g. by providing
food).
I studied direct and indirect effects of humans on several interacting species in a
terrestrial food web. First, I found that despite the minor cost of livestock predation when
evaluated in the broader context of the rural economy, it still constitutes an important
issue that compensation programs will not easily eliminate. Livestock predation will
likely continue to elicit strong direct effects of humans on wolves. Second, I found direct
effects of wolves on prey. Elk selected forest cover and low-food-quality habitat in
response to wolf presence. Cattle (Bos taurus) selected roads (likely to seek human
protection) and low-food-quality habitat, but only after wolves left pastures, indicating
poor anti-predator behaviour. Despite presence of the top-down effects described above,
path analysis indicated that humans influenced species distribution from the bottom-up
with direct effects on forage and positive indirect effects on herbivores (elk and cattle)
and wolves. As a result, human presence influences multi-species assemblages. An
ii
overall assessment of mammalian species (humans included) relative density in the study
area demonstrated that large herbivores were three times more abundant on high-use
roads and trails whereas predators were less abundant.
I documented strong direct and indirect effects of humans on an entire food web.
To fully understand and where necessary mitigate the impact of humans on species in an
ecosystem, managers must not limit their considerations to direct effects, but also
consider the indirect effects on species at several trophic levels of food webs.
iii
PREFACE
Chapter two and chapter three of this thesis were previously published in peer-reviewed
journals. I was first author on both publications and was the primary person responsible
for designing research, collecting data, analyzing data and writing the publication.
Permissions to include my previously published work in this thesis were obtained from
the journal publishers (Elsevier, License Number: 2493120632479; John Wiley and Sons,
License Number: 2493130794057). Full citations for that work are:
Muhly, T.B. and Musiani, M. 2009. Livestock depredation by wolves and the ranching
economy in the Northwestern U.S. Ecological Economics 68:2439–2450.
Muhly, T.B., Alexander, M., Boyce, M.S., Creasey, R., Hebblewhite, M., Paton, D., Pitt,
J.A. and Musiani, M. 2010. Differential risk effects of wolves on wild versus domestic
prey have consequences for conservation. Oikos 119:1243-1254.
iv
ACKNOWLEDGEMENTS
Many individuals and organizations provided significant support for this work
that was crucial to its successful completion.
I would especially like to thank my Ph.D. supervisor Dr. Marco Musiani. I am
very appreciative to him for taking me on as his first Ph.D. student. It was his unwavering
enthusiasm and generosity with his time that helped me succeed. Marco has been an
excellent mentor and will continue to be my friend.
I thank my supervisory committee: Dr. Danielle Marceau, Dr. Greg McDermid
and Dr. Paul Paquet. The success of this work is a testament to the support they provided,
especially during the formative period of my Ph.D. work. I thank Dr. Colleen Cassady St.
Clair, Dr. Mark Hebblewhite, Dr. Susan Kutz, and Dr. Judit Smits for serving as external
examiners on thesis defense and candidacy exams.
I am very thankful to my colleagues in the Faculty of Environmental Design and
the University of Calgary. I would especially like to thank Dr. Allan McDevitt, Andrew
Jakes, Byron Weckworth, Carly Sponarski, Christina Semeniuk, Isabelle Laporte, Laura
Hickman and Mathieu Pruvot. They are some of the most generous people I know and I
am very grateful for having been able to not only work with them but become friends.
I am extremely appreciative of the support and generosity of my family. I am
especially grateful to my wife Suzie, my mum and dad Linda and Bryon, my brother
Nathan and my grandparents Doris, Glen, Jean, and Mel, for their unequivocal and
unconditional love and support.
I thank the following individuals for their invaluable contributions that supported
my work: Roger Creasey for his vision, leadership and guidance in establishing and
v
overseeing the Montane Elk Project that contributed to an incredible amount of valuable
research in southwest Alberta, Dr. Mark Hebblewhite for his generosity with his time and
significant guidance on many aspects of my work, Carly Sponarski, Isabelle Laporte and
Dr. Allan McDevitt for their hard work and positive attitude in the field and in the lab
that made the work enjoyable, Dr. Carita Bergman for her vision and important
contribution to establishing the Montane Elk Project and my research, Dale Paton and
Justin Pitt for diligent planning and collection of elk telemetry data, Dr. Mark Boyce for
his guidance and wisdom, Mike Alexander for collecting and sharing cattle telemetry
data and assisting me with the planning and logistics of portions of my research, Dr.
Alessandro Massollo, Christina Semeniuk and Laura Hickman for their excellent
collaboration and important contributions to chapter five, Dr. Ed Bork for assisting me
with study design and sharing cattle telemetry data, Joe Northrup for his collaborative
spirit and for collecting traffic counter data, Elisabetta Tosoni and Suzanne Stone for
their important work collecting and organizing the livestock depredation compensation
data, Terry Mack who collected and shared wolf telemetry data, Cristina Eisenberg and
Dr. John Vucetich for discussions and positive support that helped me clarify my ideas
and significantly improve my work, Kimo Rogala and Rachelle Haddock for teaching me
how to set-up camera traps, Dr. Tak Fung for assisting with path analysis, Al Heschl and
Brian Sundberg for logistical support in the field, Richard Kennedy for support and
patience with the elk telemetry database, and several anonymous reviewers for their
insightful comments on my published work.
I would like to thank the following organizations for their important financial and
logistical support: the communities and livestock producers of Idaho, Montana, Wyoming
vi
and southwest Alberta, the Alberta Beef Producers, Alberta Ecotrust, Alberta Sustainable
Resource Development, Alberta Tourism, Parks and Recreation, Bailey Wildlife
Foundation Compensation Trust, Calgary Foundation, the Canadian Wildlife Federation,
Defenders of Wildlife, the Faculty of Environmental Design, the Glaholt Graduate
Scholarship in Environmental Design, the Institute of Sustainable Energy, Environment
and Economy, Kendall Foundation, the Montane Elk Project, the Natural Science and
Engineering Research Council of Canada, Parks Canada, Safari Club International, Shell
Canada/Royal Dutch Shell, Southern Alberta Land Trust Society, TD Friends of the
Environment, the United States Fish and Wildlife Service, the United States Department
of Agriculture-Wildlife Services, the University of Calgary, the University of Alberta,
and the Wilburforce Foundation.
vii
TABLE OF CONTENTS
ABSTRACT ............................................................................................................................ii
PREFACE ............................................................................................................................. iv
ACKNOWLEDGEMENTS ......................................................................................................... v
TABLE OF CONTENTS ........................................................................................................viii
LIST OF TABLES.................................................................................................................... x
LIST OF FIGURES .................................................................................................................xi
LIST OF ABBREVIATIONS ...................................................................................................xvi
CHAPTER ONE: INTRODUCTION ............................................................................................ 1
1.1 Direct and Indirect Effects of Humans on Food Webs ............................................. 1
1.2 Economic Motivations of a Well-known Example of Direct Effects of Humans on
Wolves............................................................................................................................. 3
1.3 Effects of Wolves on Domestic and Wild Herbivores: Predator Control Might
Result in Indirect Effects................................................................................................. 7
1.4 Predator-Mediated Versus Forage-Mediated Indirect Effects of Humans................ 9
1.5 Human Influence on Space Use by Large Mammalian Predator and Prey Species:
Implications for Ecological Communities..................................................................... 10
1.6 Thesis Objectives and Outline................................................................................. 11
CHAPTER TWO: A CLASSIC EXAMPLE OF THE DIRECT EFFECTS OF HUMANS ON OTHER
SPECIES - LIVESTOCK DEPREDATION BY WOLVES AND THE RANCHING ECONOMY IN THE
NORTHWESTERN U.S. ........................................................................................................ 14
2.1 Introduction ............................................................................................................. 14
2.2 Methods................................................................................................................... 16
2.3 Results and Discussion............................................................................................ 21
CHAPTER THREE: DIRECT EFFECTS OF WOLVES ON OTHER SPECIES IN FOOD WEBS DIFFERENTIAL RISK EFFECTS OF WOLVES ON WILD VERSUS DOMESTIC PREY.................. 43
3.1 Introduction ............................................................................................................. 43
3.2 Methods................................................................................................................... 44
3.3 Results ..................................................................................................................... 54
3.4 Discussion ............................................................................................................... 61
CHAPTER FOUR: BROADER DIRECT AND INDIRECT EFFECTS OF HUMANS - ECOSYSTEM
ENGINEERING BY HUMANS INFLUENCES THREE TROPHIC LEVELS OF A TERRESTRIAL
FOOD WEB ......................................................................................................................... 65
4.1 Introduction ............................................................................................................. 65
4.2 Methods................................................................................................................... 67
4.3 Results and Discussion............................................................................................ 74
CHAPTER FIVE: DIRECT AND INDIRECT EFFECTS OF HUMANS ON WHOLE COMMUNITIES HUMAN USE OF ROADS AND TRAILS HELPS PREY WIN THE PREDATOR-PREY SPACE RACE
........................................................................................................................................... 85
5.1 Introduction ............................................................................................................. 85
5.2 Methods................................................................................................................... 87
5.3 Results ..................................................................................................................... 92
5.4 Discussion ............................................................................................................... 98
CHAPTER SIX: CONCLUSIONS ........................................................................................... 101
viii
6.1 The Economics of Livestock Predation by Wolves and Implications for Direct
Effects of Humans on Wolves..................................................................................... 101
6.2 Wolves Directly Affect Their Prey but Such Effects are Different for Domestic
Compared with Wild Animals..................................................................................... 103
6.3 Direct and Indirect Effects of Humans Result in Ecosystem Engineering
Structuring Whole Food Webs.................................................................................... 105
6.4 Direct and Indirect Effects of Humans can Also Influence Predator-Prey
Interactions .................................................................................................................. 106
6.5 Future Research Opportunities.............................................................................. 107
REFERENCES .................................................................................................................... 110
APPENDIX A. RESOURCE SELECTION FUNCTION COEFFICIENTS AND STANDARD ERRORS FOR
ELK AND CATTLE BEFORE, DURING AND AFTER WOLF VISITS TO HOME RANGES AND
PASTURES, RESPECTIVELY, CALCULATED FROM GENERALIZED LINEAR MIXED MODELS. .. 141
APPENDIX B: HUMAN DISTRIBTUION MODEL FOR SOUTHWEST ALBERTA, CANADA ......... 145
APPENDIX C: ELK, CATTLE AND WOLF RESOURCE SELECTION FUNCTION (RSF) MODELS FOR
SOUTHWEST ALBERTA, CANADA...................................................................................... 147
APPENDIX D: ELK, CATTLE AND WOLF RESOURCE SELECTION FUNCTION MODEL
VALIDATION ..................................................................................................................... 149
APPENDIX E: FORAGE UTILIZATION ................................................................................. 151
APPENDIX F: STRUCTURAL EQUATION MODELS INDICATING SPECIES RELATIONSHIPS WITHIN
HIGH-HUMAN AND LOW-HUMAN USE WOLF HOME RANGES............................................... 153
ix
LIST OF TABLES
Table 2.1. Occurrences of deadly wolf attacks, numbers of domestic animals killed and
consumption patterns of livestock carcasses by wolves in Idaho, Montana and
Wyoming U.S. from 1987 to 2002……………………………………………………..22
Table 2.2. Estimated value of livestock losses from wolves and, as a comparison, gross
income from livestock production in Idaho, Montana and Wyoming from 1989 to 2002.
Value
of
livestock
losses
does
not
account
for
inflation…………………………...…………………………………………………....33
Table 2.3. Number and percentage of cattle and calf deaths due to predators and nonpredator causes in Idaho, Montana and Wyoming in 2005. “Other Predators” includes
wolves, grizzly bears and black bears…………………………………………………34
Table 3.1. Changes in habitat resource selection by elk and cattle before, during and after
wolf visits to home ranges and pastures, respectively, in southwest Alberta in 20042007…………………………………………………………………………………...57
x
LIST OF FIGURES
Figure 2.1. Map of the study area in the northwestern U.S. Light gray areas within Idaho,
Montana and Wyoming indicate the range of wolf populations in those states and thus
the area within which livestock depredation can occur and where depredation data was
collected. National parks are indicated as dark grey…………………………………...17
Figure 2.2. Photo of two wolves feeding on a sheep carcass. This study’s findings on
killing in excess of food requirements suggested that wolves conducted “excessive
killing” of sheep (photo credit: Stefano Mariani, 2004)……………………..………...23
Figure 2.3. Relationship between the number of domestic animals killed or injured by
wolves and compensation disbursed for the damage in Idaho, Montana and Wyoming
U.S. from 1987 to 2002. Trend in the total number of domestic animals killed and
compensation disbursed (a), and simple linear regression (with 95% confidence
intervals) between number of animals killed and compensation dollars disbursed (b)..28
Figure 2.4. Average 2007 Consumer Price Index (CPI) adjusted annual farm real estate
value (a), and annual percent change in farm real estate value (b), for Idaho, Montana
and Wyoming from 1990 to 2003. Farm real estate value (a) provided in CPI adjusted
U.S. dollars per hectare. Percent change in farm real estate value (b) is the change in
adjusted value from the previous year to the next year except 1994, which was
interpolated assuming an equal annual change between 1990 and 1994………………36
Figure 2.5. Average 2007 Consumer Price Index (CPI) adjusted August value of livestock
(cattle, steers and heifers, calves, sheep and lambs) (a), and percent change in livestock
value (b), for Idaho, Montana and Wyoming from 1989 to 2003. Value of livestock in
(a) provided in CPI adjusted U.S. dollars per hundredweight (i.e., 100 pounds; Cwt).
xi
Percent change in (b) is the change in adjusted value from the previous year to the next
year. August is the month used to determine livestock value, as that is the month with
peak depredations (Musiani et al., 2005) and thus the value at which many livestock are
compensated……………………………………………………………………………37
Figure 3.1. Map of the study area in southwest Alberta, Canada with home ranges of wolf
packs (n=4, total of 16 wolves collared) and of elk (n=10 collared), and with cattle
pastures (n=3, total of 31 cattle collared) where wolf-prey interactions were studied in
2004-2007. Home ranges were determined using a 95% kernel density estimator……45
Figure 3.2. Experimental design used to test for the effects of wolves on elk and cattle
habitat selection in southwest Alberta in 2004-2007. The treatment phase was the
period when wolves were located within buffered cattle pastures or elk home ranges,
plus a period to account for temporal precision of the wolf relocation data, whereas preand post-phases were 18 hours long (the average treatment phase length). Also
indicated are the hierarchical strata of the data controlled for including total number of
wolf visits (level-3 stratum), prey animals with radiotelemetry collars involved (level-2
stratum), and their radiotelemetry locations (level-1 stratum; see Methods)………….48
Figure 3.3. Box plots for elk selection coefficients for distance to forest cover (A), and
high-food quality habitat (B) assessed in phases before, during and after wolf visits to
elk home ranges in southwest Alberta, 2004-2007. Conditional coefficients across wolf
visits were estimated with generalized linear mixed models (GLMMs). Phases with
different coefficients (P<0.05) are marked by different letters above the box plot,
whereas box plots with same letter are not different. Also indicated are the median
xii
value (white line within the box), 25th and 75th percentiles (bounds), and 10th and 90th
percentiles (whiskers)………………………………………………………………….58
Figure 3.4. Box plots for cattle selection coefficients for distance to forest cover (A), and
high-food quality habitat (B) assessed in phases before, during and after wolf visits to
cattle pastures in southwest Alberta, 2004-2007. Conditional coefficients across wolf
visits were estimated with generalized linear mixed models (GLMMs). Phases with
different coefficients (P<0.05) are marked by different letters above the box plot,
whereas box plots with same letter are not different. Also indicated are the median
value (white line within the box), 25th and 75th percentiles (bounds), and 10th and 90th
percentiles (whiskers)………………………………………………………………….61
Figure 4.1. The spatial distribution of humans during the day (a) and night (b), wolves
during the day (c) and night (d), elk during the day (e) and night (f), cattle during the
day (g) and night (h) and forage quality and quantity (i) along roads and trails in
southwest Alberta, Canada. Human distribution is calculated from an index of human
density on roads and trails transformed into human counts collected using road counters
and trail cameras (Appendix B). Wolf (n = 14), elk (n = 62) and cattle (n = 50) day and
night distribution models were calculated from resource selection functions (RSFs)
using satellite- and GPS-telemetry data and habitat covariates measured from
geographic information system datasets (Appendix C). The forage index is the product
of the area of high-food-quality habitat types and mean of the maximum Normalized
Difference Vegetation Index within a 1 km radius of a 30-m2 pixel. For illustrative
purposes roads and trails are exaggerated with a 1-km buffer and RSF values are binned
xiii
into five geometric interval classes (i.e., each class has approximately the same number
of values and the change between intervals is consistent)……………………………..75
Figure 4.2. Structural equation model illustrating the direction and strength of
relationship between the spatial distribution of humans, wolves, elk, cattle and forage
quality and quantity and forage utilization during the daytime (sunrise to sunset) and
nighttime (sunset to sunrise)in southwest Alberta, Canada. Solid arrows indicate causal
direction of the consumer-resource interaction and line thickness is proportional to
relationship strength (path analysis β coefficient indicated). Human influences are
represented by dashed-dotted lines. Goodness of Fit Index (GFI), Adjusted Goodness of
Fit Index (AGFI), Root Mean Square Residual (RMR) and Akaike’s Information
Criterion (AIC) values of the models indicate model fit………………………………77
Figure 4.3. A sample case of wolf telemetry locations around roads with different human
counts during the daytime compared with nighttime in southwest Alberta, Canada.
Daytime locations are from sunrise to sunset, nighttime locations are sunset to sunrise.
Human count is the average number of humans on roads and trails during the day (a),
and night (b), calculated from an access index model (Appendix B). Darker pixels
indicate higher human counts………………………………………………………….81
Figure 5.1. A sample of photos taken by cameras deployed on roads and trails in
southwest Alberta, Canada during the summer of 2008. I photographed all large
mammalian species in southwest Alberta, also including: cougar (top left), wolf (top
right), moose (bottom left) and elk (bottom right).…………….………………………93
Figure 5.2. Dendrogram of the hierarchical cluster analysis of species presence/absence
data that illustrates co-occurrence of species at camera sites in southwest Alberta,
xiv
Canada during the summer of 2008. The dendrogram is scaled with the percentage of
information remaining in the analysis, where less information remaining indicates a
weaker association between species. Clusters were identified using the Ward’s linkage
method with the Euclidean distance measure..…………………..…………………….94
Figure 5.3. Co-occurrence of species at camera sites as determined by non-metric
multidimensional scaling (NMS) ordination of species counts at camera sites in
southwest Alberta, Canada during the summer of 2008. Ordinations along axis one are
indicated. Ordination was performed in PCORD version 5.17 using Sørenson’s distance
measure (McCune and Mefford 2006)…………………………………………………95
Figure 5.4. Regression tree analysis of large mammalian predator (top) and prey (bottom)
counts at camera sites in southwest Alberta, Canada during the summer of 2008. For
each of the nodes on the tree, the explanatory variable is shown with the value that best
determines the partition (i.e., the cut-off point that maximizes homogeneity within a
group). Also indicated for each node are the number of cameras in the group (count)
and the mean number of predator or prey photographs per 100 days (and standard
deviation)………………………………………………………………………………97
xv
LIST OF ABBREVIATIONS
Adjusted goodness-of-fit index: AGFI
Akaike Information Criterion: AIC
Alberta Sustainable Resource Development: ASRD
Density-mediated interaction: DMI
Density-mediated indirect interaction: DMII
Digital elevation model: DEM
Generalized linear mixed model: GLMM
Geographic Information Systems: GIS
Global-positioning-system: GPS
Goodness-of-fit index: GFI
Non-consumptive effect: NCE
Nonmetric multidimensional scaling: NMS
Normalized Difference Vegetation Index: NDVI
Resource selection function: RSF
Root mean squared residual: RMR
Structural equation modeling: SEM
Trait-mediated interaction: TMI
Trait-mediated indirect interaction: TMII
United States Department of Agriculture-Wildlife Services: USDA-WS
United States Fish and Wildlife Service: USFWS
xvi
1
CHAPTER ONE: INTRODUCTION
1.1 Direct and Indirect Effects of Humans on Food Webs
Humans are now dominating the Earth’s ecosystems (Vitousek et al. 1997).
Although humans have significantly modified ecosystems to meet their needs for tens
of thousands of years, through habitat change (Smith 2007) and by contributing to
continental-scale extinctions of megafauna through hunting (Barnosky et al. 2004),
the growing human population and demand for resources is now driving ecosystem
change at a global scale. There is significant uncertainty whether current alteration of
ecosystems to meet human needs is sustainable because of the scale and complexity
of such impacts (Foley et al. 2005). A pivotal contemporary challenge facing humans
is how to obtain sufficient ecosystem resources to meet the needs of a growing
population yet sustain the Earth’s life support systems (Turner et al. 2007).
An important goal of applied ecology is to develop knowledge useful for
maintaining functioning ecosystems (Lubchenco et al. 1991; Palmer et al. 2004). Food
webs are a fundamental ecological characteristic of ecosystems, and how “top-down”
(i.e., predators) versus “bottom-up” (i.e., resources) forces regulate food webs has
significant implications for ecosystem function (Hunter and Price 1992; Power 1992).
Thus, important ecological questions within the context of ecosystem sustainability
are how humans influence top-down versus bottom-up regulation of food webs
through direct and indirect interaction with multiple species, and what are the
consequences for ecosystem function and ultimately humans (Chapin et al. 1997;
Tylianakis et al. 2008).
Humans can have both direct and indirect effects on other species. Direct
effects occur when there are no intermediary species between two interacting species,
for example predation or herbivory interactions. Indirect effects require an
2
intermediate species (Abrams 1995), for example, herbivore prey that mediate
interactions between predators and plants.
Direct and indirect effects include density-mediated and trait-mediated
interactions (Schmitz and Suttle 2001; Schmitz et al. 2004; Preisser et al. 2005;
Preisser and Bolnick 2008). Density-mediated interactions (DMIs), also referred to as
consumptive effects (Preisser and Bolnick 2008), occur when species cause numerical
reductions in other species populations by directly killing them (Preisser et al. 2005).
Trait-mediated interactions (TMIs; Preisser et al. 2005), also called non-consumptive
effects (NCE; Preisser and Bolnick 2008) or risk effects (Creel and Christianson
2008), are direct effects manifested as changes in morphology or behaviour of the
prey (e.g., changes in habitat use, vigilance, foraging, aggregation, movement patterns
and sensitivity to environmental conditions). Schmitz et al. (1997) specified two types
of trait-mediated effects: (1) non-lethal effects, which are changes to life history or
habitat selection that do not result in mortality; and, (2) lethal effects, where changes
to life history or habitat selection result in death, for example, due to starvation. In
sum, one species can influence another species directly by hunting and killing them,
by causing them to disperse and seek refuge, or both.
Density and trait-mediated interactions can indirectly affect other species.
Density-mediated indirect interactions (DMIIs) occur when the abundance of one
species (species A) indirectly affects another (species C) by changing the abundance
of an intermediate (species B) that interacts with both species A and C (Schmitz
1998). Similarly, trait-mediated indirect interactions (TMIIs) occur when one species
modifies the way two other species interact by causing a behavioural change (Schmitz
1998; Werner and Peacor 2003). TMII occur, for example, when the presence of
predators in a community cause prey to make behavioural choices (i.e., avoid contact
3
with predators) that results in changes to herbivore foraging patterns and thus changes
to herbivore impacts on plants (Beckerman et al. 1997; Schmitz et al. 2004).
1.2 Economic Motivations of a Well-known Example of Direct Effects of Humans
on Wolves
Humans can have strong direct effects on other species. For example, humans
hunted the passenger pigeon (Ectopistes migratorius), one of the worlds most
abundant species, to extinction (Halliday 1980). Another well-documented example
of the strong direct effect that humans can have on other species is the negative effect
of humans on wolves throughout wolf range (Muhly et al. 2010a; Fritts et al. 2003).
One of the main reasons that humans negatively affect wolves is because wolves kill
livestock, and thus actions are taken to limit wolf density and distribution in livestock
production areas.
Livestock production is an important economic activity in many parts of North
America, including the Northwestern U.S. (Bedunah and Willard 1987; Sarchet
2005), yet the livestock industry is facing challenges in maintaining its economic
viability (Wuerthner 1994; Hanson et al. 2008). One particular challenge is dealing
with the costs of wolf (Canis lupus) predation on livestock (i.e., depredation), of
which the Northwestern U.S. region provides a typical example. Since 1987,
Canadian wolves have recolonized regions of Northwestern Montana. In 1995, wolves
were reintroduced in Yellowstone National Park and central Idaho, and since then
have expanded their range into contiguous areas. Many parts of the Northwestern U.S.
now frequented by wolves overlap livestock production areas and consequently
wolves have killed livestock. Livestock depredation by wolves is therefore a financial
cost of wolf reintroduction borne by livestock producers, which creates conflict
between producers, wolves and organizations involved in conservation and
4
management that can undermine wolf conservation (Niemeyer et al. 1994; NaughtonTreves et al. 2003).
Ironically, contiguous, relatively undeveloped private rangelands can provide
habitat for wildlife outside of public land (Hobbs et al. 2008). Such landscapes are
often necessary for conservation of wide-ranging wildlife species, particularly for
large carnivores such as wolves (Woodroffe and Ginsberg 1998; Carroll et al. 2003).
In actuality, livestock production may provide indirectly an important benefit for wolf
conservation –i.e. a positive externality. Externalities (see Pigou 1932 and Baumol
1972), are the positive outcomes (benefits) or negative outcomes (costs) of an
economic activity (in this particular case, livestock production) that are not reflected
in the market price of the commodity (e.g., livestock). Typically, externalities
influence individuals and groups that are not engaged in the economic activity. In this
case, livestock production may provide a benefit to wolf conservationists, which are a
large portion of the public (Kellert et al. 1996) and come from different social
contexts than those engaged in livestock production. In practice, livestock producers
maintain landscapes that are important to wolves and wolves are important to other
groups. In accordance with externality theory, these benefits are not reflected in the
market price of livestock.
Depredation can have significant monetary costs and cause emotional stress
for individual livestock producers (Bangs et al. 1998; Bangs and Shivik 2001;
Naughton-Treves et al. 2003). Several aspects of wolf depredation damage may
contribute significantly to the perception of the problem by the affected producers.
For example, one perception held by livestock producers that likely contributes to
conflict with wolves is that wolves kill livestock at a rate beyond that necessary to
supply their immediate needs for food (i.e., “surplus killing”; Gipson et al. 1998).
5
Surplus killing is characterized by an absence of (Kruuk 1972), or a low level of
(Short et al. 2002) utilization of the prey carcasses by the predator. If wolves engage
in surplus killing of livestock they could kill a number of individuals over a short
period, potentially contributing to increased financial costs to producers. In addition,
surplus killing by wolves may significantly influence opinions of livestock producers,
as well as their perception of the costs of depredation. Surplus killing by wolves has
been documented on wild prey (DelGiudice 1998), although it is not considered
common. However, surplus killing of livestock may be relatively more frequent
because of poor anti-predator behaviour in domestic animals (see chapter three).
Although wolf depredation is an important concern of livestock producers, this
concern occurs within the larger economic argument: that is, livestock producers who
lose livestock to wolves pay a cost (i.e., what could be viewed as a negative
externality) for conserving rangelands that are critical for wolf persistence. One
means to mitigate this cost is through compensation for livestock depredation.
Compensation programs are designed with the objective to help producers financially
and to reduce or eliminate animosity towards wolves by reimbursing livestock
producers for the monetary value of livestock killed by wolves (Wagner et al. 1997;
Naughton-Treves et al. 2003; Bangs et al. 2004).
Compensation has been in place in the Northwestern U.S. for the last 20 years.
However, compensation programs can be controversial as they do not necessarily
improve attitudes of livestock producers towards wolves (Naughton-Treves et al.
2003). In fact, in some cases compensation programs may have the opposite effect by
increasing the conflict between producers and agencies. For instance programs that
take a long time to reimburse producers create the impression that the agencies
providing compensation do not take the problem seriously (Fourli 1999; Montag
6
2003). Nevertheless, halting compensation is not advisable due to potential backlash
(Naughton-Treves et al. 2003) and compensation is proposed to continue in the
Northwestern U.S. in the near future (USFWS 2008). Compensation programs such as
those in the Northwestern U.S. should be assessed to ensure producers are promptly
reimbursed.
Wildlife conservation programs that employ economic tools such as
compensation must understand the economic context within which the compensated
individuals’ industry operates, as economic factors important to that industry may
ultimately influence the success of such programs. A good example of this comes
from the U.S. management of livestock depredation by coyotes (Canis latrans). For
80 years the U.S. government promoted and funded lethal control of coyotes as a
means to improve sheep production by preventing coyote depredation of sheep
(Berger 2006). Perception was that coyote depredation was driving the decline in the
sheep industry. However, governmental lethal control ultimately had no effect on the
industry; rather, rising production costs and declining commodity prices reduced or
eliminated the profitability of sheep production (Berger 2006).
In the Northwestern U.S., poor profitability from livestock production could
threaten the livelihood of all livestock producers, not just those experiencing
depredation. This may have significant implications for not only the effectiveness of
compensation programs, but for conservation of wolves on private lands. In addition,
recent demand for natural amenities (e.g., recreational opportunities and viewscapes,
sensu Hansen et al. 2002) provided by some agricultural areas has contributed to a
trend in conversion of agricultural land to rural-residential (Hansen et al. 2002;
Sengupta and Osgood 2003; Brown et al. 2005; Gosnell and Travis 2005). Conversion
of agricultural lands to other land uses can dramatically alter habitat (Theobald et al.
7
1997; Hansen et al. 2002; Mitchell, et al. 2002) in such a way that could negatively
affect wildlife, wolves included. For example, fragmentation of rangeland into smaller
land tenure parcels results in habitat loss as well as in diminished accessibility of
resources important to wildlife (Hobbs et al. 2008).
1.3 Effects of Wolves on Domestic and Wild Herbivores: Predator Control Might
Result in Indirect Effects
The direct effect of humans on wolves has implications not only for wolf
density and distribution, but also for the density and distribution of other interacting
species in the food web through indirect effects. For example, direct effects of
humans on wolves might have indirect effects on wolf prey (i.e., large mammalian
herbivores). Recent studies on large mammal food webs in protected area ecosystems
(Hebblewhite et al. 2005a; Ripple et al. 2001) indicate predators can have
disproportionately strong effects on prey and vegetation trophic levels (i.e., “trophic
cascades”; Paine 1980) that ultimately positively contribute to biodiversity (Ray et al.
2005). The influence of predators on prey are not limited to killing of prey (i.e., lethal
or density-mediated interactions) but include the risk effects (Creel and Christianson
2008) of prey avoiding predators, also called non-consumptive (e.g., Preisser and
Bolnick 2008) or trait-mediated (e.g., Abrams 1995) effects. Risk effects manifest as
changes in behaviour adopted by prey to avoid being killed, and they ultimately can
affect prey fitness (Creel and Christianson 2008, Preisser and Bolnick 2008). Risk
effects can propagate throughout food webs via indirect effects (Agrawal et al. 2001).
The effects of predators may be vital to ecosystem integrity (Estes et al. 1996) and are
therefore of significant importance to ecosystem conservation and management.
Risk effects of canids on ungulates appear strong, particularly relative to other
predators such as mountain lions (Puma concolor) (Laundré et al. 2001, Kluever et al.
8
2009). Through their interaction with ungulates, canids may influence whole
ecosystems via-indirect effects, and wolves may be a good sample case (e.g., Crête
and Manseau 1996). Wolves are the main predator of ungulates in many parts of the
world. The risk effects on elk (Cervus elaphus) are notable because of the relatively
wide distribution and overlap between these species. In addition, implications are well
documented for other components of the ecosystem, such as riparian vegetation,
beaver (Castor canadensis) and songbirds via indirect effects (Ripple et al. 2001,
Hebblewhite et al. 2005a). Management considerations for wolves, elk and their
interactions therefore have implications for conservation of the whole ecosystem.
In many ecosystems where wolves and elk co-occur (including my study area),
domestic cattle (Bos taurus) are also present and are killed by wolves. In fact,
domestic animals such as cattle are the dominant herbivore (in terms of both
numerical abundance and biomass) in many ecosystems and might play a significant
role in mediating the effects of predators on ecosystems. In addition, predation and
harassment of domestic animals by wolves creates conflict with humans in many parts
of the world (Fritts et al. 2003; see above). Resolution of such conflicts requires
adequate understanding of the ecological context of predator–prey interactions
(Ormerod 2002). In these ecosystems it might therefore be important to understand
risk effects of predators on domestic as well as wild animals. However, the risk
effects of wolves on free-ranging domesticated large ungulate livestock are relatively
unknown (but see Kluever et al. 2009).
Artificial selection has produced populations of domestic animals with
reduced potential for survival in nature (Foley et al. 1971; Eibl-Eibesfeld 1975;
Wood-Gush and Duncan 1976; Price 1984; Mignon-Grasteau et al. 2005). In domestic
livestock production, traits with economic value (e.g., faster weight gain, more wool)
9
are favoured, which diverts resources from other traits (Mignon-Grasteau et al. 2005).
As a result, domestic animals typically have smaller brains and less acute sense
organs than do their wild ancestors (Diamond 2002). Furthermore, animals that are
less fearful of humans are preferred, and therefore domestic animals tend to be much
tamer than wild animals (Lankin 1997; Gross 1998). Domestic animals may express a
lower incidence of behaviours and morphological traits that deter predators (Johnsson
et al. 2001; Mignon-Grasteau et al. 2005) due to artificial selection by humans.
1.4 Predator-Mediated Versus Forage-Mediated Indirect Effects of Humans
As suggested above, the direct effects of humans on wolves have the potential
to influence indirectly herbivore species in food webs. Humans may therefore
ultimately initiate terrestrial “trophic cascades” to the vegetation trophic level of food
webs (Pace at al. 1999). Thus, conserving and restoring predators to ecosystems has
been argued as an important mechanism to restore ecosystem structure and function
(Manning et al. 2009; Licht et al. 2010). However, this argument assumes strong
predator top-down effects across diverse terrestrial ecosystems and relies almost
exclusively on evidence from protected area ecosystems. Outside of protected areas
human activities are less restricted, allowing for many types of direct and indirect
human influences on many species in food webs, not just influences mediated by
predators.
In particular, humans also influence the flow and availability of resources to
other organisms in food webs through physical changes to the environment by acting
as ecosystem engineers or niche constructors (Smith 2007; Jones et al. 1997). For
example, agricultural activities by humans, which modify the availability of nutrients
to plants, are known to have long-lasting effects (>1,500 years) on vegetation
communities (Dambrine et al. 2007), with probable indirect effects on higher trophic
10
levels of the food web (e.g., Elmhagen and Rushton 2007). Thus, an alternative to the
“trophic cascade” hypothesis is that humans primarily influence food webs as
“ecosystem engineers” by directly influencing forage resources and indirectly
influencing higher trophic levels (i.e., herbivores and carnivores). The “trophic
cascade” versus “ecosystem engineering” effects of humans on food webs provide
contrasting hypotheses to test how humans directly and indirectly influence food
webs.
1.5 Human Influence on Space Use by Large Mammalian Predator and Prey
Species: Implications for Ecological Communities
Humans simultaneously directly influence the density and distribution of
several large mammalian species and thus have simultaneous indirect effects on other
species in food webs. For example, negative direct effects of humans on multiple
predator species might have indirect positive effects on several prey species. Humans
might therefore significantly influence the density, distribution and interactions
among species of whole ecological communities.
The outcome of predator-prey interactions can be represented as a “space
race”, where prey try to minimize and predators try to maximize spatial overlap (Sih
2005). Overall, prey must identify space where they can obtain sufficient resources to
live (e.g. food, water, cover etc.) and avoid predators (Lima and Dill 1990), as well as
habitats that might improve escape ability from predators (Heithaus et al. 2009). In
contrast, predators can use space based on the abundance of their prey, or track the
distribution of prey resources as cues for areas preferred by prey (Flaxman and Lou
2009).
In terrestrial systems, disturbance by humans, such as activity along roads
ands trails, can also influence species use of space through similar mechanisms as
11
predator-prey interactions. Hunting and culling (i.e., lethal control) of animals by
humans obviously reduces species density in an area, and the relationship between
humans and large mammals exemplifies such mechanisms. However, humans can
more subtly influence species use of space through nonlethal mechanisms, i.e., by
influencing their behaviours and spatial distribution (Frid and Dill 2002). For
example, at high levels of disturbance humans can displace wolves even in not-hunted
protected populations (e.g., Hebblewhite and Merrill 2008).
Direct lethal and nonlethal effects of humans on a species can indirectly affect
interacting predator or prey species through trophic food webs. For example, humancaused extirpation of many large mammalian predator species from much of their
worldwide range (Treves and Karanth 2003) indirectly triggered an irruption (i.e., a
rapid increase in population) of herbivore prey species in many of those areas (e.g.,
Beschta and Ripple 2009). Similarly, human displacement of predators can boost prey
density as in these areas predation risk is decreased (e.g., Hebblewhite et al. 2005a;
Ripple and Beschta 2006). Some prey species (e.g., moose [Alces alces]) even appear
to select space close to humans (e.g., roads) in areas where predator (e.g., grizzly bear
[Ursus arctos horribilis]) densities are high as a means to avoid encounters with
predators that avoid humans (Berger 2007). Such direct and indirect effects may apply
to whole predator and prey guilds if several predator species are influenced by
humans. Human influence may ultimately tip the predator-prey “space race” in favour
of prey when humans negatively affect predators, which positively effects prey.
1.6 Thesis Objectives and Outline
The primary objective of this thesis is to document direct and indirect effects
of humans on a terrestrial food web. I studied food web interactions between humans,
large mammalian predator and prey species and vegetation (wolves, elk and grassland
12
vegetation specifically) because their interactions are well documented in several
ecosystems worldwide (e.g., Crête and Manseau 1996; Ripple et al. 2001; Creel et al.
2005; Beschta and Ripple 2009), although biased towards protected areas (e.g.,
Hebblewhite et al. 2005; Berger 2007). However, results from protected areas provide
an important contrast to results from my research in a human-dominated landscape.
Furthermore, I consider interactions with domestic cattle because of their dominance
in terms of number and biomass (Beaulieu et al. 2001; Beaulieu and Bedard 2003;
Alberta Beef Producers 2010) and their role in human-wolf conflict (e.g., Muhly et al.
2010a) in the study area.
In chapter two I discuss the human-wolf conflict regarding livestock predation
by wolves. The direct negative effects of humans on wolves (e.g., lethal control) in
response to this conflict are well documented worldwide, including the study area
(Muhly et al. 2010a). I analyzed some of the costs of livestock depredation by wolves
to livestock producers relative to recent economic trends in the livestock production
industry, specifically income generated from livestock production and trends in land
and livestock value. My objective was to determine the importance of economic costs
of livestock predation in motivating lethal control of wolves.
In chapter three I test whether wolves directly influence habitat selection by
their herbivore prey. If wolves are directly influencing their prey, then humans might
indirectly influence wolf prey, mediated by their interactions with wolves. I contrast a
wild herbivores (elk) response to wolves with a domestic herbivores (cattle) response.
I discuss how the effects of predators on other species in food webs are likely
different in areas where domestic livestock are the dominant herbivore compared with
areas where wild herbivores are dominant. My objective was to determine if wolves
were influencing prey species, including domestic animals, from the top-down.
13
In chapter four I test whether humans are indirectly influencing herbivore and
vegetation species in a food web, mediated by their direct effects on wolves.
Specifically, I test whether the negative influence of humans on wolves positively
influences elk and cattle and positively influences herbaceous vegetation utilization.
However, I also test the alternative hypothesis that humans influence wolves, elk and
cattle by improving forage vegetation. My objective was to determine if humans were
influencing the food web from the top-down, bottom-up or both.
In chapter five I test whether human activity affects the spatial distribution of
several large mammalian species within an ecological community, not just wolves,
elk and cattle. I discuss what the implications might be for not only predator-prey
interactions but also other interacting species in the food web. My objective was to
determine if effects of humans occurred on multiple predator and prey species, not
just wolves, elk and cattle.
Finally, in chapter six I summarize and discuss the implications of my work. I
emphasize that humans can have manifold direct and indirect effects on several,
interacting species of ecosystems and thus the consequences of human activity are not
always easy to predict. I discuss how scientists and managers must not only consider
direct effects when studying or mitigating the influence of humans on ecosystems, but
also consider the indirect effects on species at several trophic levels of food webs.
14
CHAPTER TWO: A CLASSIC EXAMPLE OF THE DIRECT EFFECTS OF HUMANS ON
OTHER SPECIES - LIVESTOCK DEPREDATION BY WOLVES AND THE RANCHING
ECONOMY IN THE NORTHWESTERN U.S.
2.1 Introduction
A classic example of the powerful direct effects that humans can have on other
species is the direct effect that humans have on wolves. Humans have a long history
of conflict with wolves because wolves kill valued game species and livestock.
Consequently, humans have extirpated wolves from significant parts of their range
worldwide. Conflict between humans and wolves still exists in livestock production
areas, including my study area, and humans continue to influence negatively their
density and distribution there.
The purpose of this chapter is to assess some of the economic costs of
livestock depredation by wolves to livestock producers relative to recent economic
trends in the livestock production industry, specifically income generated from
livestock production and trends in land and livestock value. Furthermore, I identified
whether surplus killing of livestock by wolves was occurring to evaluate whether it
should be considered when evaluating the monetary and non-monetary costs of
livestock depredation by wolves. My approach illustrates why livestock depredation
by wolves is an important concern to livestock producers and how trends in various
assets of livestock production, of which livestock and land value may be of paramount
importance, may ultimately affect wolf conservation in agricultural areas in the
Northwestern U.S. If, livestock predation constitutes an important issue apart from its
costs, then it will not be easily eliminated by government compensation programs and
may continue to elicit strong direct effects of humans on wolves, as previously
documented in the study area (Muhly et al. 2010a).
15
I investigated patterns and trends in livestock depredation by wolves,
compensation for depredation, and livestock and land price in Idaho, Montana and
Wyoming from 1987 to 2003. This time frame represents a period under which
wolves had endangered species status and were managed by the United States Fish
and Wildlife Service (USFWS) in the Northwestern U.S. More recently wolf
management was transferred to state authorities (USFWS 2008) and then transferred
back to USFWS jurisdiction following a court injunction. I avoided analyses during
this period of management transitions.
My approach consisted of evaluating wolf depredation in the study area from
five different perspectives. First, I described the patterns of wolf depredation on
various livestock species in the Northwestern U.S. Second, I tested whether wolves
killed domestic animals in excess of their food needs to address the consequential
perception of “surplus killing” that is held by some livestock producers. Third, I
evaluated the timing of the various steps involved in financial compensation of wolf
damage and the relationship between the monetary value of domestic animals killed
and the funds actually disbursed to assess the effectiveness of compensation at
reimbursing producers. Fourth, I compared the annual monetary value of livestock
killed by wolves with annual gross income from livestock production with the
objective to place the costs of depredation within the context of economic production
of the industry. Finally, I examined trends in livestock price and land price to evaluate
the value of these two important assets in the study area. My results provide a greater
understanding of the economics of livestock depredation by wolves, which have
implications for conservation planning in areas with significant human-wolf conflicts,
in particular on private agricultural land. In the framework of externality theory, my
analysis of the practice of livestock damage compensation also constitutes an
16
important example of the costs paid by some segments of society for the benefit of
conserving wildlife.
2.2 Methods
2.2.1 Study area
The study area consisted of the northwestern states of Idaho, Montana and
Wyoming in the U.S.A. (Fig. 2.1). Parts of the study area are characterized by
temperate steppe consisting of agricultural lands and grasslands interspersed with
stands of (Populus spp.) and (Salix spp.) Forests consist of white and black spruce
(Picea spp.), subalpine fir (Abies lasiocarpa), lodgepole pine (Pinus contorta),
trembling aspen (Populus tremuloides), balsam poplar (P. balsamifera) and white
birch (Betula platyphylla). Several prey species for wolves are abundant in parts of
the study area including bison (Bison bison), moose (Alces alces), elk (Cervus
elaphus), white-tailed deer (Odocoileus virginianus), mule deer (Odocoileus
hemionus), and bighorn sheep (Ovis Canadensis). Domestic animals, particularly
livestock such as cattle, sheep, and horses are also abundant.
The region contains both developed areas (e.g. towns, agricultural lands and
managed forests) and public reserves (e.g. Wilderness Areas and National Parks
including Glacier and Yellowstone). Major economic activities in the study area
include manufacturing, mining, petroleum and natural gas extraction, and agriculture.
Important agricultural crops include potatoes, hay, barley, wheat, peas, beans, and
sugar beets. Tourism and the service industry is an emerging economic sector due to
the amenity values in the study area. Wilderness habitats that support species such as
wolves provide environmental amenities that may stimulate economic growth in the
region (Rasker and Hackman 1996). Wolf conservation in the study area may also
generate income through tourism (Duffield 1992). Although wolves were released
17
Figure 2.1. Map of the study area in the northwestern U.S. Light gray areas within
Idaho, Montana and Wyoming indicate the range of wolf populations in those states
and thus the area within which livestock depredation can occur and where depredation
data was collected. National parks are indicated as dark grey.
18
into regions where the economy was not as agriculturally based as other parts of the
study area (Bangs et al. 1998), livestock production is an important economic activity
on the private and public grazing lands where wolves settled. Furthermore, wolf range
has expanded further into livestock production areas since reintroduction and
recolonization from Canada in the mid 1990s.
2.2.2 Wolf depredation and compensation data
I analyzed USFWS and United States Department of Agriculture-Wildlife
Services (USDA-WS) depredation investigations in Idaho, Montana and Wyoming,
from January 1987 to January 2003. The non-government organization Defenders of
Wildlife collaborated with the government by funding and administering
compensation for depredation. In the study area, “confirmed” damage caused by wolf
depredation was refunded at full market value and “probable” damage was refunded
at half market value. Investigation forms included the following information for
depredation events for which wolf responsibility was “confirmed”: dates for
complaint, investigation, and payment to the livestock producer, amount
compensated, estimated time elapsed between wolf attack and investigation, livestock
species, number and age of domestic animals killed or injured, and carcass condition
(i.e., level of consumption).
Using the investigation data, I analyzed the number of cattle and sheep killed
per attack and the amount of consumption of each carcass by wolves. Instances in
which scavengers were clearly involved (i.e., as noted by the investigator) or where
depredation events were not discovered by ranchers within 24 hours, as estimated by
government officers during investigations, were not analyzed. The latter method was
designed to minimize chances for overestimation of carcass consumption by wolves
19
due to undetected scavengers or to decomposition (Miller et al. 1985; Hayes et al.
2000).
I calculated biomass consumed by wolves and compared it with total live
weight of the carcass(es) and to the weight of parts edible by wolves. In previous
studies conducted on moose (Peterson 1977), deer (Fuller 1989), caribou (Rangifer
tarandus) (Ballard et al. 1987), and Dall sheep (Ovis dalli) (Hayes et al. 1991), an
approximate proportion of 0.75 of the live weight of an ungulate was assumed edible,
or found to be edible by wolves (bones without marrow, rumen contents and hide are
typically not consumed). A few (less than 10%) of the investigation reports included
live weight of the depredated animals. When this was not provided, I assigned
depredated animals to weight classes based on their age. For cattle, animals aged >24
months were assigned 600 kg, animals aged 12-24 months were assigned 320 kg,
whereas animals aged <12 months were assigned a weight depending on monthly
developmental stages of the calf (Berg and Butterfield 1976; USDA 2003). For sheep,
all animals reported as adult were assigned 57 kg, whereas animals reported as lamb
were assigned 16 kg (Butterfield 1988; USDA 2003). Data on pack sizes were
obtained from the USFWS and data on maximum food consumption per meal by
individual wolves were gathered from the literature (approximately 8 kg; Mech 1970;
Huggard 1993). I used the above values to calculate the following indices for wolf
consumption: biomass consumed per attack (kg), biomass consumed per domestic
animal depredated (kg), and proportion of biomass consumed per attack out of total
weight of all edible parts available (%). I used the Mann-Whitney U test (Sokal and
Rohlf 2000) to compare wolf consumption indices for cattle and sheep carcasses.
I calculated elapsed time between: (a) depredation complaints and depredation
investigations to evaluate promptness of response to complaints; (b) depredation
20
events (assessed during investigations) and complaints to evaluate promptness in
response by livestock producers to the compensation program and/or producers’
ability to find losses; and, (c) investigations and mailing of compensation checks to
evaluate promptness of the compensation payment. I used the Kolmogorov-Smirnov
test with reference to Lilliefors’ probabilities (Sokal and Rohlf 2000) to determine
that the data on the yearly numbers of domestic animals killed and injured and
compensation costs were normally distributed. To test correlation among yearly
numbers I used Analysis of Variance (ANOVA) and least-squares simple linear
regression (Sokal and Rohlf 2000). All tests were two-tailed and the significance cutoff was P<0.05.
2.2.3 Estimated monetary value of livestock killed by wolves in the Northwestern
U.S.
To estimate the monetary value of cattle and sheep killed by wolves in the
Northwestern U.S. each year from 1989 to 2003 I multiplied the number of livestock
killed each year by the average livestock weight and price per kilogram for each
livestock type: cow, steers and heifers, calves, sheep and lambs (USDA 2008). I
compared monetary value of livestock killed with annual gross income generated by
cattle/calf and sheep/lamb operations as provided by the USDA (USDA 1993; USDA
1998; USDA 2004a). To further evaluate the compensation program I compared the
number of livestock killed with published compensation disbursements provided by
Defenders of Wildlife over the same period (available from
http://www.defenders.org/) using a Spearman correlation and a paired two-tailed t-test
(Sokal and Rohlf 2000).
21
2.2.4 Livestock and farm land price in Idaho, Montana and Wyoming
From the 1990s to 2003, I assessed yearly trends in livestock and farmland
prices in Idaho, Montana and Wyoming. Data on livestock prices were available for
each month from the National Agricultural Statistics Service, Agricultural Statistics
Board, USDA. For analyses of trends, I used figures from August of each year, as
August coincides with the peak of livestock depredation by wolves in the study area
(Musiani et al. 2005) and I assumed most livestock producers were compensated for
their losses at the August market price. Livestock market prices were calculated from
the sales of livestock from producers to buyers. I obtained data on farm real estate
prices for Idaho, Montana and Wyoming for 1990 and from 1994 to 2003 (USDA
1999; USDA 2004b). Farm real estate is defined as all land and buildings (including
operator dwelling) used for agricultural production (USDA 2004b). For comparative
purposes, I adjusted all annual prices (including livestock and farm real estate) to year
2007 prices using the United States Department of Labor Consumer Price Index (CPI)
inflation calculator. I calculated least-squares simple linear regressions (Sokal and
Rohlf 2000) to determine if there was a linear trend in adjusted prices for livestock
and farm real estate value and I calculated the annual percent change in adjusted
prices, and their standard deviations to evaluate the variability of prices throughout
the study period.
2.3 Results and Discussion
2.3.1 Wolves kill sheep in excess compared with other domestic animals
In Idaho, Montana and Wyoming, U.S.A., from January 1987 to January 2003,
wolves attacked cattle in 158 instances resulting in 219 individuals killed
(Mean=1.39, SD=1.07 cattle per attack; Table 2.1). There were 68 instances of attacks
on sheep (Fig. 2.2) in which wolves killed 602 individuals (Mean=8.85, SD=14.45
22
Table 2.1. Occurrences of deadly wolf attacks, numbers of domestic animals killed
and consumption patterns of livestock carcasses by wolves in Idaho, Montana and
Wyoming U.S. from 1987 to 2002.
Kilograms
Kilograms
Percent of edible
consumed per
consumed per
parts consumed per
attack, mean ± SD carcass, mean ± SD attack, mean ± SD
Cattle
Attacks
Animals
Killed
Sheep
Attacks
Animals
Killed
Dogs
Attacks
Animals
Killed
Other
Attacks
Animals
Killed
Total
158
219
87 ± 93
79 ± 89
40 ± 31
68
602
48 ± 55
8±8
18 ± 19
…
…
…
…
…
…
…
…
…
23
32
4
8
Attacks
253
Animals
Killed
861
23
Figure 2.2. Photo of two wolves feeding on a sheep carcass. This study’s findings on
killing in excess of immediate food requirements suggested that wolves conducted
“excessive killing” of sheep (photo credit: Stefano Mariani, 2004).
24
sheep per attack). The number of sheep killed per wolf attack was comparable to other
studies (3-7.6 sheep per attack; Telleria and Saez-Royuela 1989; Fritts et al. 1992;
Fico et al. 1993; Ciucci and Boitani 1998). Wolves killed more sheep than cattle per
attack (Z=9.233, P<0.001). Other domestic animals, including dogs, were killed
infrequently. During most wolf attacks on cattle, only one individual was killed (121
instances corresponding to 77% of attacks). Two cattle were depredated in 22
instances, three in nine instances, and between four and eight in six instances
(totalling 23% of attacks with more than one cattle killed). Wolves killed one sheep
per attack in 12 instances (18% of all attacks on sheep), two in 15 instances (22% of
attacks), three in five instances, and between four and 98 in 36 instances (totalling
60% of all attacks on sheep with more than two sheep killed per attack).
Biomass consumed by wolves per attack was not significantly different
regardless if the prey was cattle or sheep (Mean=87 kg, SD=93 kg and Mean=48 kg,
SD=55 kg, respectively; Z=1.384, P=0.167; Table 2.1). Thus my data show that
instigation of attacks was influenced by actual need for food, as generally assumed in
other predation studies (Holling 1965; Sjöberg 1980). However, biomass consumed
per individual animal depredated was higher when wolves attacked cattle than when
wolves attacked sheep (Mean=79 kg, SD=89 kg and Mean=8 kg, SD=8 kg,
respectively; Z=3.490, P<0.001). Relative use of edible parts consumed per attack
was higher when wolves attacked cattle than when wolves attacked sheep
(Mean=40%, SD=31% and Mean=18%, SD=19%, respectively; Z=2.037, P=0.042),
i.e., more sheep was unused or “wasted” (sensu Kruuk, 1972). Furthermore, edible
biomass of prey killed in an attack exceeded the estimated food requirements of an
average wolf pack (48 kg for 6 wolves corresponding to ≤1 cattle and ≤2 sheep) more
frequently for sheep than for cattle (Z=5.336, P<0.001). Compared with attacks on
25
cattle, a higher proportion of attacks on sheep involved killing in excess of the
immediate food requirements of an average pack living in the study area. The original
definition of "surplus killing" (Kruuk, 1972) was the killing of prey characterized by
an absence of utilization of the carcasses. According to Short et al. (2002), instances
when a predator consumes only a small portion of the total edible prey clearly also
indicates surplus killing. Despite the inconsistencies in the definition of surplus
killing, my findings on killing in excess of immediate food requirements demonstrate
that wolves conduct “excessive killing” (sensu Carbyn, 1983; Miller et al., 1985) on
sheep.
Although my data indicate wolves can kill sheep in excess of food
requirements, further exploration of excessive killing on domestic animals might be
necessary for two reasons. Firstly, I was unable to assess whether wolves would
return to carcasses to consume more meat from livestock kills if human disturbance of
the carcass had not occurred. Human disturbance might have happened when the
rancher detected the kill and when ranchers and government personnel conducted the
subsequent investigation. Human disturbance may bias my results towards concluding
that excessive or surplus killing occurred when it would have not. However, humans
occur throughout wolf range in the study area and arguably throughout wolf range
worldwide, with a few exceptions (Fritts et al. 2003). Similar circumstances are found
for various carnivore species worldwide (Treves and Karanth 2003). Thus, human
disturbance of prey consumption could be considered as a constant part of the
environment; and this applies to both agricultural (as described above) and protected
areas.
A second methodological bias could be due to my inability to determine the
exact number of wolves that fed on an animal. As lone wolves or only a portion of the
26
pack living in the area may have been involved, my results (which account for
potential visits by a whole pack) may underestimate the total edible biomass
consumed per individual. Thus, again, my results could potentially be biased toward
overestimating the occurrence of excessive killing. Nevertheless, in the absence of
intensive experimental manipulation, more accurate or unbiased assessments of
excessive killing are not feasible.
My data highlight the greater vulnerability of sheep to predators compared
with cattle (Hell 1993; Gipson and Paul 1994; Ciucci and Boitani 1998; Keeling and
Gonyou 2001) and support the importance of prey traits (Roberts 1996; Lima 1998;
Brown 1999; Mitchell and Lima 2002) in influencing the outcome of wolf-livestock
interactions. Livestock traits such as morphology, choice of habitat, vigilance and
overall ability to defend against or escape predator attacks are influenced by animal
husbandry and artificial selection (Foley et al. 1971; Eibl-Eibesfeld 1975; Wood-Gush
and Duncan 1976; Price 1984; Johnsson et al. 2001; Mignon-Grasteau et al. 2005).
Sheep are easy prey for canids due to their small body size, low agility and lack of
defence mechanisms (i.e., horns; Short et al. 2002). Flocking behaviour, which is
naturally exhibited by sheep and reinforced through husbandry (Keeling and Gonyou
2001) can supply large numbers of highly concentrated, vulnerable prey during a wolf
attack. Confinement of sheep at high densities in enclosures permeable to wolves but
not sheep might also compel wolves to kill many sheep in a single attack (Kruuk
1972; Stuart 1986). In addition, all breeds of sheep group more closely together than
cattle (Grandin and Deesing 1998). My results suggest that differences in
morphology, behaviour and husbandry between sheep and cattle induce different
predatory behaviours by wolves, represented as excessive killing of sheep in relation
to food needs of wolves.
27
Organizations involved in wolf conservation and/or providing depredation
compensation should be aware of excessive killing of sheep by wolves. There is
potential for added negative perception of wolves by sheep producers due to excessive
killing. Excessive killing could result in a number of sheep killed by wolves in single
attacks, with the parallel need to refund quickly the damage of such sudden losses.
Compensation providers should therefore identify and regularly communicate with
sheep producers within wolf range to ensure depredation events are identified and
compensation is delivered promptly. In addition, when anticipating future costs of
depredation it should be acknowledged that sheep might be killed in excess of what
would be predicted based on normal rates of meat consumption by wolves.
2.3.2 Timing in compensation of livestock losses suggests prompt compensation
In many parts of Europe and North America a typical means to prevent or
reduce animosity of agricultural communities towards wolves is to compensate
livestock producers for damage caused by wolves (Boitani 1982; Gunson 1983; Mech
1995). I detected a strong relationship between yearly numbers of domestic animals
killed and injured by wolves and compensation costs disbursed for damage
throughout the study (linear regression, R2=0.86, F=86.85, P<0.001; Fig. 2.3). On
average, U.S. wildlife officials investigated instances of wolf depredation the same
day that complaints were received (Mean=0, SD=1 days; range=0-8 days). Most
livestock producers contacted authorities between one and two days of estimated
occurrence of death or injury to domestic animals (Mean=1, SD=3 days; range=0-21
days). Compensation checks were sent to producers between 10 and 273 days after
investigations, but on average within 77 days (SD=53 days).
My data indicate compensation is delivered promptly in the study area.
Producers appear to patrol and check their properties and livestock intensively, and
28
(a)
(b)
Figure 2.3. Relationship between the number of domestic animals killed or injured by
wolves and compensation disbursed for the damage in Idaho, Montana and Wyoming
U.S. from 1987 to 2002. Trend in the total number of domestic animals killed and
compensation disbursed (a), and simple linear regression (with 95% confidence
intervals) between number of animals killed and compensation dollars disbursed (b).
29
government investigations are timely. However, some losses might be undetected by
ranchers (Oakleaf et al. 2003) or detected late when carcass decompositions preclude
identification of the offending predator species (Bangs and Shivik 2001). Thus, my
sample, which deals only with confirmed losses due to wolves, likely overestimates
promptness of detection. I found the major delay in the compensation program was in
the sending of the actual compensation check -within an average of 2-3 months.
However, the timing of refunding was comparable or shorter than other compensation
programs for carnivore damage (Fourli 1999). Even so, compensation programs could
be improved by reducing the delay in delivering compensation money. In general,
prompt payment is believed to contribute to ease public animosity (Wagner et al.
1997; Fourli 1999). Producers invest in livestock every year and they benefit from
weight gain and reproduction. Any delays in compensation result in additional losses
to the producers, because of lack of such growth and reproduction in the
compensation-delay period. Prompt compensation can reduce such additional losses.
Furthermore, any delays in compensation would suggest to producers that the
agencies providing compensation do not take the depredation problem seriously
(Montag 2003), and thus delays could undermine the effectiveness of compensation at
reducing animosity towards wolves.
The positive influence of compensation programs on livestock producers’
tolerance for wolves or other depredating carnivores is debatable (Naughton-Treves et
al. 2003; Gusset et al. 2009). Ongoing wolf control in the study area (Bangs and
Shivik 2001; Bangs et al. 2004; USFWS 2008) suggests that compensation has not
necessarily improved tolerance of livestock producers toward wolves. There are
several potential problems with compensation, including for example, disparity in the
political views of agencies and livestock producers over who is responsible for
30
damage, differences in urban and rural values and inability to address larger issues of
large carnivore conservation, for example, human safety concerns (see Montag 2003
for review). In addition, many producers believe compensation is inadequate because
it does not compensate for emotional investment in livestock and because many wolf
kills go uncompensated due to lack of detection or definitive evidence (NaughtonTreves et al. 2003). Furthermore, compensation does not typically account for
“welfare loss” of depredation (Bostedt and Grahn 2008), i.e., the value of an animal
outside of its meat value; for example, the financial loss if the killed animal is an
important breeding animal. Nevertheless, the benefits of ongoing wolf damage
compensation programs appear to outweigh their costs and there cancellation is not
advocated (Chambers and Whitehead 2003; Naughton-Treves et al. 2003). Rather,
organizations currently involved in compensation programs should be aware of the
limitations of such programs for improving tolerance for wolves and enhancing wolf
conservation.
2.3.3 Costs of livestock depredation by wolves appear minimal at the industry scale,
although significant at the ranch scale
According to calculations based upon confirmed kills and average market
price for livestock, wolves killed an average of $11,076.49 worth of livestock per year
between 1987 and 2003, with a maximum of $41,230.32 in 2003 (Table 2.2). The
estimated costs of livestock depredation increased during this period (linear
regression, R2=0.789, slope=2445.161, P<0.001; Table 2.2). My estimates were
correlated with published compensation disbursements provided by Defenders of
Wildlife over the same period (spearman correlation, r=0.986, P<0.001); however, my
depredation cost estimates were lower than costs disbursed (t=-3.101, P=0.008).
These findings perhaps indicate willingness also to refund un-confirmed losses in
Year
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
Cattle
$1,632.21
$3,257.72
$1,192.86
$585.80
$0.00
$3,174.66
$1,355.84
$4,590.34
$10,124.73
$8,986.59
$14,667.43
$15,243.39
$20,833.68
$22,211.57
$34,593.68
Sheep
$0.00
$0.00
$47.72
$0.00
$0.00
$0.00
$0.00
$1,315.99
$5,753.82
$398.98
$3,321.93
$3,087.73
$5,269.56
$2,921.34
$6,636.64
Livestock losses from
wolves
Cattle and calf operations
income (Gross, x 1,000)
$2,041,756.00
$1,920,076.00
$1,976,956.00
$2,035,723.00
$2,110,752.00
$1,990,894.00
$1,820,334.00
$1,688,947.00
$2,096,127.00
$2,018,890.00
$2,109,207.00
$2,469,288.00
$2,662,758.00
$2,540,105.00
NA
Sheep and lamb operations income
(Gross, x 1,000)
$66,696.00
$53,096.00
$55,639.00
$71,234.00
$73,988.00
$68,986.00
$85,548.00
$82,024.00
$88,774.00
$70,384.00
$67,290.00
$71,068.00
$56,717.00
$61,230.00
NA
Wyoming from 1989 to 2002. Value of livestock losses does not account for inflation.
Cattle and calf losses
relative to income
0.0001%
0.0002%
0.0001%
<0.0001%
<0.0001%
0.0002%
0.0001%
0.0003%
0.0005%
0.0004%
0.0007%
0.0006%
0.0008%
0.0009%
NA
Sheep and lamb losses
relative to income
<0.0001%
<0.0001%
0.0001%
<0.0001%
<0.0001%
<0.0001%
<0.0001%
0.0016%
0.0065%
0.0006%
0.0049%
0.0043%
0.0093%
0.0048%
NA
Table 2.2. Estimated value of livestock losses from wolves and, as a comparison, gross income from livestock production in Idaho, Montana and
31
32
some circumstance, with the likely aim to foster tolerance for wolves. I estimated an
annual monetary value of cattle killed by wolves of $9,496.70. Sheep monetary value was
approximately six times smaller ($1,579.79). However, the proportion of monetary loss
relative to gross income of the industry in Idaho, Montana and Wyoming was one order
of magnitude smaller for cattle than for sheep (Table 2.2); thus, sheep producers
experienced proportionally higher costs due to wolves. In either case, the monetary value
of cattle and sheep killed by wolves was less than 0.01% of the total income of all
producers.
To give further perspective to the depredation problem, I analyzed data gathered
in Idaho, Montana and Wyoming in 2005 which documents the relative contribution of
predators to cattle mortality from all causes. The number of cattle and calf losses due to
the category called “other predators” (including wolves, grizzly bears and black bears)
was no more than 3% of all mortality (Table 2.3). This result suggested that within those
states with wolf populations, the total costs of cattle depredation by wolves to the
industry is relatively low compared with the costs of other losses (i.e., total non-predator
losses accounted for at least 89% of all losses). Unfortunately, such data were not
available for sheep, nor does it describe the relative contribution of wolves compared
with other predators. Although, I have no reason to believe that losses due to wolves to
other livestock species would be substantially different.
Despite the relatively low cost of wolf depredation to the livestock industry,
depredation could be comparatively costly at the ranch scale. In particular, individual
producers that experience multiple depredations (Niemeyer et al. 1994; Naughton-Treves
et al. 2003) or excessive killing (see above) may carry a higher burden of the costs. In
33
Table 2.3. Number and percentage of cattle and calf deaths due to predators and nonpredator causes in Idaho, Montana and Wyoming in 2005. “Other Predators” includes
wolves, grizzly bears and black bears.
Coyotes
Dogs
Mountain
Lions and
Bobcats
Other
Predators
Total
Predator
Digestive
Problems
Respiratory
Problems
Metabolic
Problems
Mastitis
Other
Diseases
Cattle
% Total Cattle
Calves
% Total Calves
Cattle
% Total Cattle
Calves
% Total Calves
Cattle
% Total Cattle
Calves
% Total Calves
Cattle
% Total Cattle
Calves
% Total Calves
Cattle
% Total Cattle
Calves
% Total Calves
Cattle
% Total Cattle
Calves
% Total Calves
Cattle
% Total Cattle
Calves
% Total Calves
Cattle
% Total Cattle
Calves
% Total Calves
Cattle
% Total Cattle
Calves
% Total Calves
Cattle
% Total Cattle
Calves
% Total Calves
Idaho
600
1
100
<1
200
0
400
1
600
1
400
1
1,500
2
3,200
8
20,000
32
6,900
16
19,600
31
1,800
4
1,500
4
2,100
5
2,500
4
State
Montana
1,300
3
100
<1
200
0
200
1
500
1
200
1
2,100
5
1,500
8
5,500
12
3,100
16
7,800
17
200
1
300
1
2,200
11
1,100
2
Wyoming
100
1
2,100
7
100
<1
100
1
400
1
300
3
900
3
500
5
3,500
11
1,200
11
5,500
18
2,700
25
6,000
19
200
2
600
5
800
3
34
Lameness/
Injury
Weather
Related
Calving
Problems
Poisoning
Theft
Other NonPredator
Unknown
Non-Predator
Total NonPredator
Total
Cattle
% Total Cattle
Calves
% Total Calves
Cattle
% Total Cattle
Calves
% Total Calves
Cattle
% Total Cattle
Calves
% Total Calves
Cattle
% Total Cattle
Calves
% Total Calves
Cattle
% Total Cattle
Calves
% Total Calves
Cattle
% Total Cattle
Calves
% Total Calves
Cattle
% Total Cattle
Calves
% Total Calves
Cattle
% Total Cattle
Calves
% Total Calves
Cattle
Calves
2,600
6
500
1
200
<1
1,100
2
3,300
8
5,900
9
500
1
300
<1
100
<1
100
<1
2,000
5
1,000
2
17,300
41
10,000
16
41,500
99
61,000
98
41,900
62,500
1,000
5
600
1
1,200
6
10,200
22
1,500
8
11,300
25
800
4
300
1
300
1
3,500
18
900
2
4,200
22
5,300
12
19,200
99
43,600
95
19,400
45,700
1,000
9
6,000
19
800
7
7,000
23
800
7
700
2
100
1
500
2
3,100
28
1,000
3
10,500
95
27,500
89
11,000
31,000
35
other words, the costs of depredation are not evenly distributed across the industry
(Collinge 2008). A few attacks with significant losses of livestock may disproportionately
influence intolerance for wolves, with negative effects for conservation. Research on risk
perception suggests that people focus on maximal events rather than the average
(Lehmkuhler et al. 2007); therefore, despite low costs to the industry, livestock producers
within wolf range may focus on and be concerned about high cost of some depredation
events, such as those documented in this study. Finally, costs of depredation are
increasing (Table 2.2), which also likely concerns producers.
2.3.4 Increasing land price and decreasing cattle price may negatively affect livestock
production and wolf conservation
Land price increased in the Northwestern U.S. during the study period (Fig. 2.4).
There was a significant increase in CPI adjusted average annual farm real estate for Idaho
(linear regression, R2=0.954, slope=88.887, P<0.001), Montana (linear regression,
R2=0.876, slope=16.262, P<0.001), and Wyoming (linear regression, R2=0.943,
slope=19.490, P<0.001), and all three states together (linear regression, R2=0.966,
slope=41.638, P<0.001). Annual rates of change were mostly positive since 1994 (Fig.
2.4b) with a mean value of 2.96% (SD=1.61%), 2.20% (SD=2.70%), 3.09% (SD=1.25%)
and 2.81% (SD=1.13%) for Idaho, Montana and Wyoming, and all three states,
respectively. The rate of change was negative in only two years (1997 and 1999), and
only in Montana. Thus, land prices were generally increasing at a steady and predictable
rate during the study period.
The price of cattle decreased during the study period (Fig. 2.5). There was a
significant negative trend in the price of cows (linear regression, R2=0.749, slope=-8.538,
36
(a)
Land value (CPI adjusted $/hectare)
4000
3500
3000
2500
ID
MT
2000
WY
Average (ID, MT, WY)
1500
1000
500
0
03
20
02
20
01
20
00
20
99
19
98
19
97
19
96
19
95
19
94
19
93
19
92
19
91
19
90
19
89
19
Year
(b)
7
6
Land value change (%)
5
4
3
ID
MT
2
WY
Average (ID, MT, WY)
1
0
0
20
0
20
0
20
0
20
9
19
9
19
9
19
9
19
9
19
9
19
9
19
3
2
1
0
9
8
7
6
5
4
3
-1
-2
-3
Year
Figure 2.4. Average 2007 Consumer Price Index (CPI) adjusted annual farm real estate
value (a), and annual percent change in farm real estate value (b), for Idaho, Montana and
Wyoming from 1990 to 2003. Farm real estate value (a) provided in CPI adjusted U.S.
dollars per hectare. Percent change in farm real estate value (b) is the change in adjusted
value from the previous year to the next year except 1994, which was interpolated
assuming an equal annual change between 1990 and 1994.
37
(a)
600
Livestock value (CPI adjusted $/Cwt)
500
400
Cow
Steers and Heifers
300
Calves
Sheep
Lambs
200
100
0
03
20
02
20
01
20
00
20
99
19
98
19
97
19
96
19
95
19
94
19
93
19
92
19
91
19
90
19
89
19
88
19
Year
(b)
60
20
Cow
Steers and Heifers
0
Calves
03
20
02
20
01
20
00
20
99
19
98
19
97
19
96
19
95
19
94
19
93
19
92
19
91
19
90
19
89
19
Livestock value change (%)
40
Sheep
Lambs
-20
-40
-60
Year
Figure 2.5. Average 2007 Consumer Price Index (CPI) adjusted August value of livestock
(cattle, steers and heifers, calves, sheep and lambs) (a), and percent change in livestock
value (b), for Idaho, Montana and Wyoming from 1989 to 2003. Value of livestock in (a)
provided in CPI adjusted U.S. dollars per hundredweight (i.e., 100 pounds; Cwt). Percent
change in (b) is the change in adjusted value from the previous year to the next year.
August is the month used to determine livestock value, as that is the month with peak
depredations (Musiani et al. 2005) and thus the value at which many livestock are
compensated.
38
P<0.001), steers and heifers (R2=0.436, slope=-6.024, P=0.007), and calves (linear
regression, R2=0.268, slope=-7.823, P=0.048). There was no significant trend in sheep
and lamb price. The average annual percent change in livestock price was -2.08%
(SD=12.63%), -0.03% (SD=13.18%), 0.92% (SD=20.90%), -0.67% (SD=17.55%), and 0.24% (SD=18.16%), for cows, steers and heifers, calves, sheep, and lambs, respectively.
Contrary to land prices (Fig. 2.4b), changes in livestock prices varied between positive
and negative values across years (Fig. 2.5b) and livestock prices were much more
variable than land prices (i.e., they had a much larger standard deviation relative to the
mean). These findings indicate volatility in the price of livestock.
Decreasing (in the case of cattle) or stagnant (in the case of sheep) trends in
livestock prices coupled with increasing land prices might be of concern to groups
interested in conservation of wildlife on or adjacent to agricultural lands. Throughout the
U.S. and including Idaho, Montana and Wyoming, demand for agricultural land that
provides natural amenities is increasing (Hansen et al. 2002; Talbert et al. 2007). In fact,
agricultural lands within or adjacent to wildlife (including wolves) and wildlife habitat
are in even greater demand and thus command higher prices (Duffield 1992; Rasker and
Hackman 1996; Bastian et al. 2002). This demand for amenities has resulted in
subdivision of agricultural ranchlands in favour of rural-residential development
(Sengupta and Osgood 2003; Gosnell and Travis 2005).
2.3.5 Implications for wolf conservation planning
I documented excessive killing of livestock for sheep, but not for cattle, and thus
could confirm the perception of livestock producers that wolves may kill some types of
livestock in excess of food needs. Fear that such behaviour could occur has contributed to
39
negative attitudes towards wolves (Gipson et al. 1998). Thus, managers and scientists
should acknowledge that wolves can conduct excessive killing of sheep, as they are
highly vulnerable due to their morphology, husbandry and behaviour. A single
depredation event may have a greater impact on the welfare of an individual sheep
producer than cattle producer, which should be accounted for in compensation programs
so that those farmers experiencing excessive killing are compensated promptly and
efficiently.
In the period during which wolves had endangered species status and were
managed by the USFWS with compensation provided by Defenders of Wildlife, the only
noteworthy delay in the compensation program occurred at the delivery of the payment
(2-3 months); however, timing was comparable to other programs around the world.
Nevertheless, reduction in this delay could potentially improve the program by allowing
producers to purchase quickly new livestock so they can profit from weight gain and
reproduction. My estimates of the monetary value of cattle and sheep killed were
correlated with, albeit significantly lower than the published amount of compensation
disbursed by Defenders of Wildlife. Perhaps the discrepancy reflected willingness by
compensation providers to refund livestock producers also for unconfirmed loses in the
hopes of improving tolerance for wolves. However, it is unclear whether higher levels of
compensation translate to increased tolerance (see Naughton-Treves et al. 2003).
Compensation for livestock depredation by wolves might help those livestock
producers that experience depredation offset some of the financial burden for conserving
wolves; however, compensation programs as they are currently designed are inadequate
for recognizing positive externalities of livestock production to wolf conservation. Other
40
economic tools can recognize these externalities and thus perhaps further encourage
tolerance for wolves and improve wolf conservation on agricultural lands.
Various studies highlight the importance of agricultural rangelands for wildlife
conservation in Western North America (Fritts et al. 1994; Hobbs et al. 2008). Some
authors object that in these areas wildlife conservation also presents challenges (see
Wuerthner 1994). For example, excessive livestock grazing can degrade habitat for wild
herbivores (Milton et al. 1994) and thus potentially reduce the abundance of wild prey for
wolves. In any case, conversion to rural-residential developments may be a worse
alternative as it will dramatically change landscape structure by increasing human
presence, as well as fragmenting and removing habitat (Theobald et al. 1997; Mitchell et
al. 2002; Sengupta and Osgood 2003). The implications of rangeland subdivision may be
particularly negative for wide ranging herbivores, and their predators (including wolves),
as habitat fragmentation can restrict access to forage, water and other resources on the
landscape (Hobbs et al. 2008). Therefore, livestock production takes place in agricultural
rangelands that provide positive externalities (sensu Pigou 1932; Baumol 1972) in the
form of large tracts of heterogeneous wildlife habitat on private land. These externalities
are recognized as clear benefits by some conservationists.
Given the economic trends I identified above, conversion of agricultural land to
rural-residential development appears to be a realistic possibility in the Northwestern
U.S. I suggest that if wolf conservation is a recognized societal objective (Musiani and
Paquet 2004), then public funds destined to wolves may be used to contribute to habitat
conservation more directly –i.e. not just to refund wolf damage to livestock. In practice,
wolf conservation programs could contribute funds to support ecologically friendly
41
livestock production in intact rangelands. Availability of such funds could be justified to
wolf conservationists by explaining the benefits offered to large sectors of society by
such areas.
One example of an economic tool that could be used to more effectively achieve
wolf conservation in agricultural rangelands are conservation performance-payments,
which are monetary or in-kind payments made by an agency to individual or groups of
landowners that achieve specific conservation outcomes (Albers and Ferraro 2006; Zabel
and Holm-Müller 2008). In the case of wolf conservation in the Northwestern U.S., for
example, payments could be made to livestock producers that conserve a certain number
of wolves on their land; producers that conserve more wolves would receive larger
payments. Conservation performance-payments may therefore provide incentive to
conserve wolves. Lethal control of wolves would theoretically become a less-desirable
response to depredation for livestock producers, as payments would decrease with fewer
wolves. The program would also put fewer burdens on livestock producers to find and
report livestock depredation events, and the need to promptly investigate and reimburse
losses would no longer be required by the compensating agency (Zabel and Holm-Müller
2008). Furthermore, such programs might encourage livestock producers to actively
adopt and deploy husbandry practices that prevent depredation, as opposed to current
programs that compensate for killed animals. Currently, many compensation programs
“reward” livestock producers whose animals are killed, rather than producers who
prevent depredation, and thus do not encourage producers to prevent livestock
depredation. Ultimately, compensation payments could also help livestock producers
maintain their operations despite poor livestock prices. I caution, however, that the
42
outcomes of conservation performance-payment programs are still uncertain and
therefore should not be widely applied without further testing of their effectiveness.
Nevertheless, the economics of wolf conservation may become more explicit with
conservation performance-payments, as the costs of wolves to livestock producers and
the benefits of agricultural rangelands for wolf conservation are defined by the payment.
The approaches explained above may not resolve the conflicts around wolves,
because such conflicts may arise from societal issues of broader relevance. The array of
people and groups interested in wolves or affected by wolf presence may be in conflict
for reasons other than wolves. Thus, surrogate conflicts may arise that are indicative of
broader disputes (not directly wolf-related). For example, people coming from urban
areas may be in conflict with rural residents, people relying on academic knowledge may
experience tension with those relying on local knowledge, and people relying on
“traditional principles” might not appreciate “modern values” or their promoters. In the
study area, livestock producers might oppose wolf reintroduction and conservation due to
negative feelings prompted as a means to protest federal agencies from establishing
greater regulation on private landowners (Montag 2003; Naughton-Treves et al. 2003).
Direct negative effects of humans on wolves may remain strong where wolf range
overlaps with livestock production unless such issues are resolved.
43
CHAPTER THREE: DIRECT EFFECTS OF WOLVES ON OTHER SPECIES IN FOOD WEBS DIFFERENTIAL RISK EFFECTS OF WOLVES ON WILD VERSUS DOMESTIC PREY
3.1 Introduction
In chapter two I discussed issues regarding livestock predation by wolves (Canis
lupus) in livestock production areas. Such issues can ultimately determine wolf density in
a region, a phenomenon observed in the study area (Muhly et al. 2010a) as well as in
other areas throughout wolf distribution (Fritts et al. 2003). The subsequent question is
whether absence of wolves determines changes in other ecosystem components, with the
obvious candidates being wolf prey.
Risk effects of predation are a function of the hunting mode of the predator, antipredator tactics of the prey and the landscape (Heithaus et al. 2009). Disentangling the
contribution of landscape from anti-predator behaviours in risk effects is challenging.
Wolves, elk and cattle would be ideal species to compare risk effects of a predator on
domestic and wild herbivores. In fact, elk responses to wolves are often dependent on
landscape features such as security cover provided by terrain (Frair et al. 2005), forest
(Fortin et al. 2005) and humans (Hebblewhite et al. 2005a). However, the direction of
such risk effects might be difficult to predict. For example, elk might select open habitats
so they can detect wolves from afar (Hebblewhite et al. 2005b) or they might select forest
cover to hide from wolves (Creel et al. 2005). Similar habitat-mediated risk effects might
be evident in cattle, or they might be absent due to artificial selection by humans that deemphasizes anti-predator behaviours. The risk effects of predators on domestic animals
may have different indirect effects on other species in ecosystems (e.g., vegetation)
compared with risk effects on wild prey.
44
The purpose of this chapter was to test for and compare risk effects of wolves, as
measured by habitat selection of wild (elk; Cervus elaphus) and domestic (cattle; Bos
taurus) ungulate species. Elk and cattle are the dominant herbivores in the study area, and
both are preyed upon by wolves. My objectives were to determine whether (1) the
presence of wolves within elk home ranges and cattle pastures caused changes in
selection of security cover and food-quality habitat patches, and (2) such risk effects were
different in cattle compared with elk. For elk, my hypotheses was that wolves would have
a significant and immediate risk effect, as the majority of sources report elk selection for
security cover and a switch to habitats characterized by sub-optimal food quality during
wolf presence. Conversely, I hypothesized that cattle would not switch habitat during
wolf presence periods, because of poor anti-predator behaviour. I compared risk effects
of satellite-collared wolves on habitat selection by global-positioning-system (GPS)collared elk and cattle in southwest Alberta, Canada by using generalized linear mixed
models (GLMMs) to calculate resource selection functions (RSFs) in periods before,
during and after wolf visits in elk home ranges or cattle pastures.
3.2 Methods
3.2.1 Study area
The study occurred within a montane ecosystem along the eastern slopes of the
Rocky Mountains in southwest Alberta, Canada (Fig. 3.1). The continental divide bounds
the western edge of the study area. Towards the east, topography is less rugged, with
rolling foothills that eventually level to flat prairie and agricultural lands. Forested lands
occur in the western half of the study area and open into grasslands to the east. Cattle
(Bos taurus) were the dominant domestic herbivore and numbered in the 10,000 to
45
Canada
Figure 3.1. Map of the study area in southwest Alberta, Canada with home ranges of wolf
packs (n=4, total of 16 wolves collared) and of elk (n=10 collared), and with cattle
pastures (n=3, total of 31 cattle collared) where wolf-prey interactions were studied in
2004-2007. Home ranges were determined using a 95% kernel density estimator.
46
100,000’s (Alberta Beef Producers 2010), but domestic sheep (Ovis aries) also occur in a
few areas. The study area encompassed four wolf pack home ranges (Laporte et al. 2010),
51 grizzly bears (ASRD and ACA 2010), >1,000 elk ([Cervus elaphus] ASRD 2002) and
an unknown number of cougars (Puma concolor), black bears (Ursus americanus),
coyotes (Canis latrans), white-tailed deer (Odocoileus virginianus) and mule deer (O.
hemionus). Wolves are the predominant predator of elk (Hebblewhite et al. 2005b) and
cattle (Musiani et al. 2003) in the region.
There are several small towns (populations of 300 to 3,665 people) located within
the study area. Agriculture (primarily livestock grazing), forestry, natural gas
development and recreational activities are the prevailing human land uses. Lands to the
west are predominantly public, and to the east are predominantly private. Livestock
grazing occurs on private lands throughout the year and on public lands in the summer
(May to October).
3.2.2 Wolf, elk and cattle telemetry data
Sixteen wolves from four known wolf packs in the study area (Fig. 3.1) were
fitted with ARGOS satellite-radiotelemetry collars (LOTEK Ltd, Aurora, ON) in
southwest Alberta between 2004 and 2007 by Alberta Sustainable Resource Development
(ASRD), using either modified foot-hold traps or helicopter netgunning. There was at
least one ARGOS collar maintained in each pack during the study. ARGOS collars were
programmed to provide three locations per day. In all wolf visits to elk and cattle no more
than one wolf was monitored and different individuals were monitored in the same pack
in different years. Therefore, the absence of a radiotelemetry collared wolf does not
necessarily mean an absence of wolves, because other wolves in the study area might
47
have been present although un-collared. Thus, the analyses that follow are conservative
because the pre- and post-wolf phases may be contaminated with some wolf presence.
However, if a prey response signal is found, this has to rise above the background ‘noise’
of possible presence of other predators throughout the study.
Forty-eight female elk were captured in January 2007 using helicopter netgunning
and fitted with either a GPS4400L or GPS2200L (LOTEK Ltd, Aurora, Ontario) GPSradiotelemetry collar programmed with a two-hour relocation schedule. On average 1,111
locations were used to create home ranges and mean elk home range size was 112 km2.
Cattle GPS-radiotelemetry data were collected over three summers (2004-2006) from
cattle in three large fenced cattle pastures (12 km2, 23 km2 and 23 km2) within the study
area (Fig. 3.1). In 2004, GPS-radiotelemetry collars (GPS3300L, LOTEK Ltd, Aurora,
Ontario) programmed with a 20-minute relocation schedule were placed on nine
randomly chosen cattle from July 1st to September 14th. In 2005, eleven cattle were
fitted with the same collars from April 1st to May 1st and from July 1st to September
10th. The same sampling design was employed in 2006, with the exception that no
radiotelemetry collars were deployed from April 1st to May 1st. Cattle collared were
yearling males and females.
3.2.3 Study design
I compared habitat selection by elk and cattle between three phases (Fig. 3.2):
before wolf visit (pre-phase), during wolf visit (treatment phase) and after wolf visit
(post-phase). A post-treatment phase was included as there can be a lag in prey responses
to predators (McGarigal and Cushman 2002). I used radiotelemetry data to document
wolf visits. Cattle with radiotelemetry collars were confined to one of three pastures (Fig.
48
Cattle
Elk
Legend
Pre-phase wolf location
Treatment phase wolf location
Post-phase wolf location
Pasture
Before
(prephase)
Cattle
18 hours
Elk
18 hours
Home
Range
During
(treatment
phase)
Time wolf in
pasture plus 4.5
hours before
and after
Time wolf in
home range
plus 4.5 hours
before and after
After
(postphase)
Number
of Wolf
Visits
(Level-3)
Number
of
Collared
Animals
(Level-2)
Number of
Radiotelemetry
Locations
(Level-1)
18 hours
19
31
20,961
18 hours
51
10
1,908
Figure 3.2. Experimental design used to test for the effects of wolves on elk and cattle
habitat selection in southwest Alberta in 2004-2007. The treatment phase was the period
when wolves were located within buffered cattle pastures or elk home ranges, plus a
period to account for temporal precision of the wolf relocation data, whereas pre- and
post-phases were 18 hours long (the average treatment phase length). Also indicated are
the hierarchical strata of the data controlled for including total number of wolf visits
(level-3 stratum), prey animals with radiotelemetry collars involved (level-2 stratum), and
their radiotelemetry locations (level-1 stratum; see Methods).
49
3.1, Fig. 3.2). I categorized wolves as present (i.e., the treatment phase) when a collared
wolf occurred within the cattle pasture or a 1.5 kilometre buffer around the pasture (Fig.
3.2). It should be considered that wolves often travel in packs and that prey species are
adapted to detect presence of packs through scent, hearing and visual clues. A detection
distance of one to two kilometres has also been assumed to identify short-term predation
risk response by prey in other studies of large mammal predator-prey interactions (Creel
et al. 2005; Gude et al. 2006). I included this buffer as a conservative means to account
for detection distance of predators by prey. I identified wolf presence periods by their
presence in elk home ranges or cattle pastures, rather than proximity to collared animals,
because elk and cattle live in herds and thus the behaviour of uncollared animals using
other parts of the home range or pasture could influence the behaviour of the collared
animals.
For elk, which range freely throughout the study area, wolf presence was
considered when a wolf with a radiotelemetry collar entered an elk home range (Fig. 3.2).
I estimated a winter and a summer home range for each radiotelemetry collared elk, as
elk in the Rocky Mountains have different seasonal home ranges (Hebblewhite et al.
2006). Winter was defined as January 12th (when the study began) to May 31st and
summer as June 1st to October 13th (when the study ended), comparable to previous
studies on elk (Creel et al. 2005; Hebblewhite et al. 2006). I estimated elk home ranges
using a 95% kernel density estimator (Seaman and Powell 1996) with a smoothing
parameter (h=3,000 m) determined based on my knowledge of elk distribution in the
study area. I did not buffer the home ranges because visual inspection of the kernels
50
indicated that the 95% contour extended 1 to 2 km outside the actual telemetry locations,
a length comparable to the buffer I used around cattle pastures.
Although more temporally resolved than previous predation risk studies (e.g.,
Creel et al. 2005), the temporal precision of my wolf relocation data (one location every
nine hours) made it difficult to determine exactly when a wolf entered and left elk home
ranges and cattle pastures. As a conservative means to account for this uncertainty I
added 4.5 hours (i.e., half of the average duration between locations) before and 4.5 hours
after the time wolves occurred in the elk home range or cattle pasture. A uniform pre- and
post-phase duration was used to ensure experimental phases were consistent and thus
comparable to each other. Pre- and post-phases were defined as 18-hour periods before
and after the treatment phase, because 18-hours was the average length of all treatment
phases. Furthermore, an 18-hour period was appropriate to minimize the effects of daily
patterns in resource selection on the results. Patterns in habitat selection typically occur
over a 24-hour (i.e., daily) cycle (Alcock 2005). These daily patterns occur in wolves
(Merrill and Mech 2003) and elk (Godvik et al. 2009) and may occur in cattle too. Thus,
tests comparing behaviours occurring in shorter periods may show significant differences
that are simply due to changes in daily patterns in activity –i.e. not due to other
‘treatments’ such as for example a predators’ visit.
3.2.4 Habitat selection by elk and cattle
I used RSFs (Manly et al. 2002; Johnson et al. 2006; Lele 2009) to quantify elk
and cattle habitat selection before, during and after wolf visits. I followed RSF “sampling
protocol A” (Manly et al. 2002) where used (i.e., elk and cattle radiotelemetry locations)
and available resource units were sampled (Johnson et al. 2006). Available resource units
51
were sampled at one random location/km2 in each home range (for elk) or pasture (for
cattle). Thus for this analysis, I considered resource selection within the home range
(third order of selection; Johnson 1980). This approach may underestimate rare habitats
within the home range that are used by elk and cattle as refugia from prey, therefore my
approach is conservative for detecting habitat switches by prey in response to predator
presence.
I considered spatial covariates for elk and cattle RSF calculation and measured the
following at each used and available location: distance to nearest road or trail, distance to
nearest forest cover, terrain ruggedness and food quality (high or low). Covariates were
quantified using Geographic Information Systems (GIS) datasets. I identified three
measures of security cover (human activity, forest and topography). The use of human
activity as a measure of security for animals might seem paradoxical. However, wolves
are known to avoid humans more than elk, which may use human activity areas as a
refuge from predation (Hebblewhite et al. 2005a). A similar difference can be
hypothesized between wolves and cattle, which might seek humans for security cover. I
used a GIS dataset of roads and trails obtained from Alberta Sustainable Resource
Development (ASRD). I defined forest security cover for ungulates as forested areas with
a canopy closure above 75% (Lyon 1979; Skovlin et al. 2002) and using a GIS model of
canopy closure (McDermid et al. 2009) I calculated the distance of each location to forest
cover with 75% canopy closure. I calculated terrain ruggedness (Riley et al. 1999) from a
30-m2 spatial resolution digital elevation model (DEM). I obtained a 30-m2 spatial
resolution GIS map of vegetation cover derived from Landsat data (McDermid et al.
2009) and collapsed it into two ungulate food quality classes, high and low. Herbaceous,
52
shrub, regenerating cutblocks, and deciduous forest cover types were defined as highfood quality and coniferous forest and barren ground cover types were defined as lowfood quality. I identified high-food quality habitats based on the preferred food types of
elk and cattle where they co-occur (Stewart et al. 2002; Beck and Peek 2005). Overall,
classification accuracy of the vegetation cover map was 80%, as calculated from a
ground-truthing approach using 245 independent, randomly selected test sites surveyed in
the field (McDermid et al. 2009). All GIS work was conducted in ArcGIS 9.2 (ESRI
Inc.). Habitat present in elk home ranges (n=10) had the following characteristics: mean
terrain ruggedness index = 13, mean distance from roads and trails = 394 m, mean
distance from forest cover = 157 m, proportion of high-food quality habitat = 83%.
Habitat present in the three cattle pastures was characterized by mean terrain ruggedness
index = 16, mean distance from roads and trails = 245 m, mean distance from forest
cover = 42 m, proportion of high-food quality habitat = 60%.
I also tested for interactions between proximity to security cover and high-food
quality habitat made by ungulates when wolves were present compared with when they
were absent. I tested for the following three interactions as covariates in the RSF:
distance to road or trail×food quality, distance to forest cover×food quality and terrain
ruggedness×food quality. A significant interaction coefficient in the RSF implies that
selection for high-food quality habitat is a function of proximity to security cover.
To test my hypotheses for elk and cattle response to wolves, I produced one RSF
model for each herbivore species and for each experimental phase that included all
covariates. I tested the strength of covariates in each RSF using a z-test (StataCorp 2009)
and examined whether there was a clear change in the sign (positive or negative) and
53
significance of coefficients when wolves were present compared with when they were
absent.
I validated each of my RSF models using k-fold cross validation (Boyce et al.
2002) at the population level (i.e., including all animals and visits). The k-fold cross
validation procedure was performed five times withholding 20% of the data at each
iteration. The area-adjusted frequency of animal locations was compared with the
predicted RSF scores using a Spearman rank correlation. A predictive model has a
significant positive correlation.
3.2.5 Hierarchical modeling of habitat selection
I used GLMMs to include random effects in my RSFs because they account for
autocorrelation of longitudinal telemetry data, correlation within hierarchical strata of the
data and between-strata variance, which allows for robust inference and appropriate
estimation of marginal (i.e., population) and conditional (i.e., visit) responses (Skrondal
and Rabe-Hesketh 2004; Gillies et al. 2006). GLMMs are increasingly used in ecology
(Bolker et al. 2009) and RSF modelling (Koper and Manseau 2009) for these reasons.
The hierarchical strata of the data (Fig. 3.2) were radiotelemetry locations (level-1)
within individual animals (level-2) within wolf visits (level-3). I therefore used a threelevel GLMM to model cattle habitat selection and included: (1) a random intercept at the
individual level (level-2) to accommodate variation in sample size of telemetry locations
(level-1) among individual cattle; (2) a random intercept at the visit level (level-3) to
accommodate for unbalanced sample size of GPS-radiocollared cattle (level-2) between
visits; and, (3) a random coefficient at the visit (level-3) to examine conditional
variability in prey use of resources in the presence of predators. I used a two-level
54
GLMM to model elk habitat selection and included: (1) a random intercept at the visit
level (level-3) to accommodate unbalanced sample size of telemetry locations between
visits; and, (2) a random coefficient at the visit level (level-3) to examine conditional
variability in prey use of resources in the presence of predators. Compared with cattle, a
random intercept at the individual level was not necessary in the elk RSFs because only
one elk was involved in each visit and thus there was no level-2 stratum.
Two GLMMS were calculated for each experimental phase, one with a
conditional level-3 coefficient for distance to forest cover and one with a conditional
level-3 coefficient for food habitat quality. I calculated the conditional value of these
covariates specifically because my initial analysis found that they were significant fixed
effects in both the elk and cattle GLMMs and that the coefficients differed between
phases of a visit (i.e., before, during and after wolf visits) indicating they were important
habitat covariates to prey in the presence of predators. I tested whether there was a
difference in the conditional coefficients for distance to forest cover or food habitat
quality between each experimental phase using a paired sample Wilcoxon signed-ranks
test (Sokal and Rholf 2000). GLMMs were calculated using STATA 10.1 (StataCorp
2009) and the GLLAMM function with a logit link (www.gllamm.org; Skrondal and
Rabe-Hesketh 2004). I derived maximum-likelihood estimates using adaptive quadrature
with 20 integration points (Rabe-Hesketh et al. 2005).
3.3 Results
3.3.1 Occurrence of wolf visits to elk home ranges and cattle pastures
I documented 51 independent wolf visits (i.e., presence events) to elk home
ranges that involved ten different elk (Fig. 3.2). Wolf visits (i.e., the treatment phase)
55
were initiated at various periods of the day. Twenty-nine visits were initiated between
12:01 AM and 6:00 AM, six were initiated between 6:01 AM and 12:00 PM, ten were
initiated between 12:01 PM and 6:00 PM and six were initiated between 6:01 PM and
12:00 AM. I also documented 19 wolf visits to cattle pastures, involving 31 different
cattle and these visits were also initiated throughout the day. Seven visits were initiated
between 12:01 AM and 6:00 AM, three were initiated between 6:01 AM and 12:00 PM,
five were initiated between 12:01 PM and 6:00 PM and four were initiated between 6:01
PM and 12:00 AM.
3.3.2 Changes in habitat selection by elk
I found a change in selection of high-food quality habitat and security cover by
elk when wolves were present in elk home ranges compared with when they were absent.
Elk avoided forest cover prior to wolf visits (z=2.970, P<0.01). According to the
marginal RSF model, before wolf presence, elk had a 95% probability of selecting areas
up to 600 m distant from forest cover (all other covariates at their mean value). Elk
neither selected nor avoided forest cover during and after wolf presence suggesting that
they used forest cover as it was available: average distance 157 m (Table 3.1; I direct
readers to Appendix A, Tables A1 and A2 for actual RSF coefficient values for elk and
cattle GLMMs, respectively). The conditional (individual treatment) coefficients for
distance to forest cover dropped during wolf visits compared with before wolf visits (Fig.
3.3A; z=7.082, P<0.001) and increased after wolf visits compared with during wolf visits
(Fig. 3.3A; z=-3.196, P<0.01). There was also a lag effect, because conditional
coefficients were lower after wolf visits compared with before wolf visits (Fig. 3.3A;
z=7.027, P<0.001). Elk selected high-food quality habitat before (z=1.993, P<0.05) and
56
after (z=2.182, P<0.05) wolf visits, but not during wolf visits (Table 3.1). According to
the marginal RSF model, before wolf presence elk had a 70% probability of occurring in
high-food quality habitat vs. 30% in low quality. During wolf presence there was neither
selection nor avoidance of high-food quality habitat suggesting elk occurred in high-food
quality habitat as it was available. As observed in the period ‘before’, after wolf presence
elk had again a higher probability of occurring in high-food quality habitat (84%).
Similar to what I found for forest cover, the conditional coefficients for high-food quality
habitat dropped during wolf visits compared with before wolf visits (Fig. 3.3B, z=4.454,
P<0.001), indicating high-food quality habitats were less important to elk when wolves
were present. There was also a lag effect, as conditional coefficient values were
significantly lower after wolf visits compared with before wolf visits (Fig. 3.3B, z=5.338,
P<0.001). Out of all the interactions tested between environmental covariates, the
coefficient for distance to road or trail and food was the only one that changed across
experimental phases and it switched from being not significant to negative during (z=2.010, P<0.05) and after (z=-2.746, P<0.01) wolf visits (Table 3.1). This indicated that
when wolves were present, and after wolves left elk home ranges (i.e., a lag effect), elk
preferred high-food quality habitat close to roads and trails. In all phases, the coefficients
of the interaction between distance to forest cover and food was negative (before: z=5.366, P<0.001, during: z=-1.789, P=0.074, after: z=-2.824, P<0.01) and between terrain
ruggedness and food was positive (before: z=2.071, P<0.05, during: z=3.634, P<0.001,
after: z=2.335, P<0.05) with similar trade-off mechanisms, which in this case did not
change with wolf visits (Table 3.1). This indicated that elk were more likely to select
57
Table 3.1. Changes in habitat resource selection by elk and cattle before, during and after
wolf visits to home ranges and pastures, respectively, in southwest Alberta in 2004-2007.
Resource
Distance to road or trail
Before
NS
Elk
During
NS
Distance to forest cover
High-food quality habitat
Distance to road or trail x
High-food quality habitat
+*
+*
NS
NS
NS
+*
+***
+***
+***
+***
+***
─***
NS
─*
─*
─***
─***
+***
Distance to forest cover x
High-food quality habitat
─***
─**
─***
─***
─***
─***
Terrain ruggedness x
High-food quality habitat
+*
+***
+*
+***
+***
+***
* = significant P < 0.05
** = significant P < 0.01
After
NS
Before
+***
Cattle
During
+***
After
─***
*** = significant P < 0.001
NS = P > 0.05
58
(A)
a
b
c
(B)
b
a
b
Figure 3.3. Box plots for elk selection coefficients for distance to forest cover (A), and
high-food quality habitat (B) assessed in phases before, during and after wolf visits to elk
home ranges in southwest Alberta, 2004-2007. Conditional coefficients across wolf visits
were estimated with generalized linear mixed models (GLMMs). Phases with different
coefficients (P<0.05) are marked by different letters above the box plot, whereas box
plots with same letter are not different. Also indicated are the median value (white line
within the box), 25th and 75th percentiles (bounds), and 10th and 90th percentiles
(whiskers).
59
high-food quality habitat when close to the security provided by forest cover and terrain
ruggedness.
3.3.3 Changes in habitat selection by cattle
Cattle response to wolves was detectable in the period after wolf visits, consistent
with a lagged effect. Cattle selected for roads and trails after wolves left pastures (z=12.555, P<0.001), whereas before (z=8.633, P<0.001) and during (z=8.113, P<0.001)
wolf visits they avoided them (Table 3.1). According to the marginal RSF model, before
wolf presence, cattle had a 95% probability of selecting areas up to 1,200 m from a road
or trail. During wolf presence, cattle had a 95% probability of selecting areas just up to
900 m from a road or trail. This distance further dropped in the period after wolf presence
when cattle had a 95% probability of selecting areas right on roads and trails. Similarly,
high-food quality habitat was selected before (z=8.401, P<0.001) and during (z=11.324,
P<0.001) wolf visits, but avoided after wolves left pastures (z=-12.721, P<0.001; Table
3.1; Fig. 3.4B). According to the marginal RSF model, before and during wolf presence
cattle had a 65% and 75% probability (respectively) of selecting high-food quality
habitat. However, after wolf presence the probability of selecting high-food quality
habitat dropped to 37%. Cattle always avoided forest cover and did so incrementally
before, during and after wolf visits (Table 3.1; Fig. 3.4A). The interaction between food
and distance to forest cover (before: z=-7.451, P<0.001, during: z=-6.355, P<0.001, after:
z=-6.974, P<0.001) and food and terrain ruggedness (before: z=11.364, P<0.001, during:
z=8.524, P<0.001, after: z=7.172, P<0.001) were negative and positive across all phases,
respectively (Table 3.1). Similar to elk, cattle preferred high-food quality habitat close to
forest cover and in more rugged terrain. The interaction between high-food quality
60
(A)
b
b
a
(B)
a
a
b
Figure 3.4. Box plots for cattle selection coefficients for distance to forest cover (A), and
high-food quality habitat (B) assessed in phases before, during and after wolf visits to
cattle pastures in southwest Alberta, 2004-2007. Conditional coefficients across wolf
visits were estimated with generalized linear mixed models (GLMMs). Phases with
different coefficients (P<0.05) are marked by different letters above the box plot,
whereas box plots with same letter are not different. Also indicated are the median value
(white line within the box), 25th and 75th percentiles (bounds), and 10th and 90th
percentiles (whiskers).
61
habitat and distance to roads and trails switched from negative before (z=-11.222,
P<0.001) and during (z=-1.972, P<0.05) wolf visits to positive after (z=15.223, P<0.001)
wolf visits. Cattle preferred high-food quality habitat close to roads and trails prior to and
during wolf visits, but switched to avoiding high-food quality habitat close to roads or
trails after wolf visits.
3.3.4 Elk and cattle RSF model validation
Elk RSFs during and after wolf visits were valid for population level inferences
(rs=0.76, P=0.018 and rs=0.75, P=0.028, respectively), consistent with the notion that
habitat selection was predictable when influenced by predation risk. The elk RSF before
wolf visits did not validate (rs=0.67, P=0.071) due to lack of correlation in one of the test
groups (rs=0.37, P=0.285). The variability of the relationship between elk presence and
habitat variables that are important in anti-predator strategies, as exemplified by lack of
validation of my RSF, might also be due to random presence of uncollared wolves as well
as other predators in the period before the wolf visits detected in this study. The cattle
RSFs were valid for population level inferences in all periods (rs=0.991, P<0.001,
rs=0.922, P<0.001 and rs=0.752, P=0.018, before, during and after wolf visits,
respectively).
3.4 Discussion
I identified a habitat-mediated risk effect of wolf presence on elk selection for
forest cover and avoidance of quality food patches. Elk immediately responded to wolf
visits, and there was a persistence of the response immediately after wolves left the home
range. My results were consistent with the notion broadly presented in other studies in
North America that elk increase their selection of forest cover and decrease selection of
62
high-food quality habitat types in the presence of predators (Morgantini and Hudson
1985; Creel et al. 2005; Fortin et al. 2005). I also found that after wolves visited home
ranges elk might have used human activity as refugia from predation. Such behaviour is
not unusual in wild ungulates (Berger 2007). Human activity has been implicated in
creating a low predation risk scenario for non-migratory elk in a similar area of Alberta
(Hebblewhite et al. 2005a).
Similar to elk, cattle made trade-offs between forest cover and food quality of the
habitat patches used. This suggested that cattle might not have been able to select for the
best food patch available and the best available forest cover scenario at the same time.
Such trade-offs presented whether wolves were present or absent. However, unlike elk,
when wolves entered pastures cattle did not respond by selecting for forest cover. It is
possible that the range of habitats available to cattle to select from during wolf visits was
limited by fencing –i.e. cattle could use habitats within pastures only. However, cattle
pastures were of comparable size to elk home ranges (Fig. 3.1) and encompassed a
similar array of habitats.
Wolves caused cattle to switch to low-food quality habitat and use roads and trails
after wolves had left the pastures. Cattle selection for areas farther from forest was a
constant tendency across phases that become stronger during and after wolf visits. Thus,
instead of forest, cattle selected roads and trails indicating they perceived human
infrastructure and activity as a form of security cover from predation. Cattle may have
remained in open areas during predator presence because they relied on grouping as an
anti-predator strategy (Laporte et al. 2010), like some large wild bovids (Hunter and
Skinner 1997).
63
The lagged responses to wolf visits indicated cattle responded less promptly than
elk to wolves, and suggested that anti-predator behaviours in cattle could have been
blunted by domestication and artificial selection. Alternatively, the lack of an immediate
anti-predator response might be due to lack of experience, instead of being the result of
artificial selection and domestication. As suggested in other behavioural studies
conducted on cattle (Kluever et al. 2009) and sheep (Romeyer and Bouissou 1992),
artificial selection has likely resulted in attenuated anti-predator behaviours in livestock
that might make them more vulnerable than wild ancestors to predation.
As is typical in ecological studies, my results might have been confounded by a
number of variables that could not be measured or controlled for in complex
environments, and that can still affect results. For example, snow accumulation might
contribute to the outcome of predator-prey interaction on certain days (Nelson and Mech
1986). Similarly, age of prey might be a factor, as younger animals might be more
vulnerable (Husseman et al. 2003). Sample sizes for my study animals and events, and
the data available on the ecosystems were all limited, which precluded from controlling
for a number of factors and variables. Therefore, my findings are limited to correlative
patterns between certain variables or factors (example, occurrence of a visit by collared
wolves) and other variables or factors (example, a change in habitat selection by prey).
Overall, I found a clear signal of a change in habitat selection by prey correlated to visits
by collared wolves, despite a conservative methodology designed to minimize Type I
errors, i.e., identifying a prey response when one did not exist.
Finally, my methodologies did not account for the possibility of wolf responses to
elk and cattle behaviour and habitat use. In theory, it is possible that wolves approach
64
prey animals only when these move to certain habitats, instead of prey responding to
wolves by moving in these same habitats. However, I made the reasonable assumption
that wolves had not detected the exact location of individual prey animals prior to
entering the elk home ranges or cattle pastures, and thus the type of habitat “used” by
prey did not influence a wolf’s decision to enter the home range or pasture. My
assumption seems supported by my finding that in the period after wolf visits various
animals (elk) returned to habitat used prior to wolf visits. Thus, it seems that a prey
response was elicited by the ‘wolf-treatment’, rather than vice-versa.
The possibility of indirect effects of humans on large mammalian herbivores,
mediated by their interaction with wolves, exists in the study area. Humans have negative
direct effects on wolves, spurred by issues regarding livestock predation, and wolves
influence elk and to some degree cattle habitat selection.
65
CHAPTER FOUR: BROADER DIRECT AND INDIRECT EFFECTS OF HUMANS - ECOSYSTEM
ENGINEERING BY HUMANS INFLUENCES THREE TROPHIC LEVELS OF A TERRESTRIAL
FOOD WEB
4.1 Introduction
In chapter three I documented the direct effects of wolves on elk and cattle habitat
selection. Given that humans can have severe direct effects on wolves (chapter two), in
theory humans may have strong indirect effects on herbivores, mediated by wolves. Such
indirect effects could also influence other ecosystem components, particularly vegetation
species. In this chapter I test the hypothesis that humans exert indirect effects on
herbivore and vegetation species in food webs, mediated by their interaction with wolves
(i.e. the “trophic cascade” hypothesis). However, I contrast this with an important
alternative hypothesis that humans indirectly influence large mammalian species through
their direct effects on forage (i.e., the “ecosystem engineering” hypothesis).
The “trophic cascade” and “ecosystem engineering” concepts provide alternative
hypotheses to test how humans influence food webs and ultimately ecosystem structure
and function, i.e., do humans primarily influence ecosystems through predators or by
modifying resources? To test these hypotheses I studied how humans concurrently
influenced the distribution of several large mammalian species across a terrestrial food
web in an ecosystem that was not a protected area (i.e., southwest Alberta, Canada). My
study species included wolves (Canis lupus), elk (Cervus elaphus), cattle (Bos taurus)
and grassland vegetation because they are focal management indicators in the region and
are well studied in terrestrial food web ecology (Johnson 2010). Wolves were the
predominant predator of elk (Hebblewhite et al. 2005b) and cattle (Musiani et al. 2003),
66
and elk and cattle are known to modify their behaviour in response to wolves (Laporte et
al. 2010; Muhly et al. 2010b; chapter three) and drive bottom-up (e.g., herbivory)
processes (Hobbs 1996). Cattle were the dominant herbivore in the region in terms of
numbers and biomass, and livestock production was a major industry in the study area
(Alberta Beef Producers 2010). Furthermore, wolves (e.g. Hebblewhite and Merrill 2008;
Musiani et al. 2010) and elk (Preisler et al. 2006) are known to respond to human
presence, and conflict between humans and wolves exists in the study area (Musiani et al.
2003). Finally, grassland vegetation is integral to wild and domestic herbivores in the
region (Morgantini and Hudson 1989).
I used path analysis, a form of structural equation modelling (SEM) (Kelloway
1998) to test and contrast the trophic cascade and ecosystem engineering hypotheses.
Specifically, path analysis was used to model the strength and direction of direct and
indirect relationships between the spatial distributions of humans (measured using
counter data and a spatial human density index) wolves, elk and cattle (measured using
telemetry data and resource selection functions [RSFs]), forage quality and quantity
(hereafter, forage; measured using remotely sensed vegetation type and biomass data) and
forage utilization (measured at plots in the field) as predicted by each hypothesis. I tested
my hypotheses within different areas and time periods characterized as having different
human density treatments, including: (1) day (high-human density period) versus night
(low-human density period), and (2) two wolf home ranges, one with the highest and the
other with the lowest human density in the study area.
If humans were primarily acting as ecosystem engineers, I predicted a positive
pathway from humans to forage, forage to herbivores (i.e., elk and domestic cattle),
67
herbivores to wolves and forage to wolves. Alternatively, if humans were causing
trophic-cascades I predicted a negative pathway from humans to wolves, wolves to
herbivores (both wild and domestic) and wolves to forage utilization, and a positive
pathway from herbivores to forage utilization. My main assumptions were that trophic
relationships between species in my food web could be captured by measuring their
spatial overlap (e.g., Peterson and Robins 2003; McLoughlin et al. 2009) and that human
counts on roads and trails (i.e., density) were correlated with all types of human land use
intensity (i.e., habitat change) (Ellis and Ramankutty 2008).
My results demonstrated that humans primarily had indirect effects on the food
web by modifying forage resources, supporting the ecosystem-engineering hypothesis.
Human impacts on forage productivity might cause food webs to switch from predator to
forage regulated. Studying direct and indirect effects of humans on multiple trophic
levels of food webs is necessary for effective ecosystem restoration and conservation.
4.2 Methods
4.2.1 Study Area
For a full description of the study area, please refer to chapter three, section 3.2.1.
4.2.2 Spatial Models of Human Distribution
To measure the spatial distribution of humans in the study area I reproduced Apps
et al.’s (2004) spatial index of human density in a Geographic Information System (GIS).
I also collected counts of humans using trail cameras (n=43) and traffic counters (n=43)
deployed in the study area in 2008. Trail cameras (RECONYX Silent ImageTM Model
RM30) were deployed using a randomized design, on roads and trails within wolf and elk
home ranges in the study area. Trail cameras provided time-stamped photographs of
68
humans (vehicles, ATV’s, hikers, etc.) that passed in front of the camera’s infrared
sensor. Pneumatic road tube counters (Diamond Traffic Products), which provide a time
stamp when a vehicle drives over the tube, were deployed non-randomly on targeted
roads to ensure spatially comprehensive counts of humans throughout the study area.
Daytime and nighttime counts of humans were statistically different at all road
and trail counter locations (n=86, daytime mean=1,030±3,284 SD and nighttime mean =
332±1,087 SD; z=8.054, P<0.0001; tested using a Wilcoxon Matched-Pairs Test).
Wildlife, including wolves (Hebblewhite and Merrill 2008) and elk (Schultz and Bailey
1978), might change their distribution in response to these temporal changes in human
distribution. Therefore I produced separate spatial models of human distribution for the
daytime and nighttime. Daytime was defined as sunrise to sunset (5:36 to 19:18) and
nighttime as sunset to sunrise (19:19 to 5:35) (National Research Council of Canada
2010).
To test the fit of the human index to actual counts of humans I modeled the
average daily and nightly counts of humans at trail camera and traffic counter locations
with the calculated index of human density in a linear regression. I had a significant
linear fit (Appendix B, Fig. B1), which suggested that the index was a good predictor of
actual human counts. Therefore I transformed the index into estimated counts of humans
during daytime and nighttime across the study area.
4.2.3 Models of Wolf, Elk and Cattle Distribution
4.2.3.1 Using Telemetry Technology to Obtain Location Data
I obtained satellite and GPS telemetry location data from wolves, elk and cattle in
the study area to model each species’ distribution during the day and night. Data were
69
obtained from wolves fitted with ARGOS satellite telemetry collars between 2004 and
2007 by Alberta Sustainable Resource Development (ASRD), captured using either
modified foot-hold traps or helicopter net-gunning. ARGOS collars were programmed to
provide a location every nine hours. In total I obtained 7,462 locations from 14 wolves (5
adult males, 3 young males, 4 adult females, and 2 young males), with a spatial precision
of ≤1 km according to the ARGOS Location Quality Index. Elk location data were
obtained using GPS or ARGOS satellite telemetry collars (Lotek 4400 and 2200 series
and Telonics TGW-3600) between January 2007 and March 2009. Elk were captured in
the winter from seven elk herds in the region, using helicopter net-gunning. GPS and
ARGOS collars placed on elk were programmed to provide a location every two hours
(spatial accuracy +-5 m; Laporte 2008). I obtained 267,440 locations from 62 elk (36
females and 26 males). Finally, cattle location data were obtained from GPS telemetry
collars (Lotek 3300L) deployed on cattle within two separate grazing allotments in the
study area. In the first allotment, data were collected from 40 different yearling steers
(males) and heifers (females) over three years (2004-2006). Cattle were collared from 1
July to 14 September. Collars were programmed with a 20-minute relocation schedule. In
the second allotment, collars were deployed on 10 adult females in 2007, from 25 June to
12 October. In total I obtained 348,514 locations from 50 cattle (spatial accuracy +-37 m;
Laporte 2008). All cattle were randomly selected from herds and captured and collared
by ranch managers. The wolf, elk and cattle capturing, handling and monitoring protocols
for this research were reviewed and approved by the Universities of Alberta and Calgary
and by the Alberta Government (Permit Numbers: BI-2008-19, RC-06SW-001 and
23181CN).
70
4.2.3.2 Using Resource Selection Functions to Model Species Distributions
I used the telemetry location data from each species to develop RSFs to
characterize the selection of habitats by each species and the resultant predicted
distribution of the species on the landscape. RSFs are any function that is proportional to
the probability of selection of a resource unit by an animal (Manly et al. 2002). Thus they
can be used to model species distributions in a GIS by contrasting resource units selected
by a species and the distribution of habitat attributes across the landscape. My objective
was to estimate a population-level RSF from a sample of individual animals of each
species. I therefore followed the two-stage method for calculating population-level RSFs
(Fieberg et al. 2010), where an RSF is calculated for each individual animal and the
individual’s RSF coefficient values are averaged across all individuals to obtain a
population average (e.g., Sawyer et al. 2006). For each species’ RSF I included habitat
covariates that previously have been used to model wolf, elk and cattle habitat selection.
Human counts were significantly different between daytime and nighttime and elk and
wolves might respond to those differences (see above), therefore I calculated daytime and
nighttime population-averaged RSFs for each species (Appendix C, Table C1).
Population-level RSFs of wolves, elk and cattle were predictive of each species’
distribution according to cross-validation methods (Appendix D, Table D1). Regression
statistical analyses were completed using STATA 10.1 (2009).
4.2.4 Distribution of Forage Quality and Quantity
I created a spatial model of the distribution of forage for herbivores. First, I
collapsed a 16-category vegetation cover GIS dataset (McDermid et al. 2009) into two
forage food quality categories: high and low. The high-food quality category consisted of
71
herbaceous, shrub and deciduous forest cover types, which are more preferred by
ungulates, whereas the low-food quality class consisted of coniferous forest, mixed forest
and bare ground cover types less preferred by ungulates (Muhly et al. 2010b; chapter
three). I calculated the area of high-food-quality habitat within a 1-km radius of each 30m2 pixel on the landscape. Second, I obtained a 250-m2 spatial resolution GIS data set of
the maximum Normalized Difference Vegetation Index (NDVI) measured during the
2005 growing season. NDVI is an index of vegetation biomass that is useful to monitor
the effect of vegetation on animal’s at large scales (Petorelli et al. 2005). I calculated the
mean of the maximum NDVI values within a 1 km radius each 30-m2 pixel. I multiplied
the maximum NDVI value by the area of high-food quality habitat to obtain an index of
forage quality and quantity.
4.2.5 Forage Utilization
I measured forage utilization by herbivores in 2007 and 2008 at 150 plots
deployed using a randomized design within the herbaceous vegetation cover type and
within the extent of wolf and elk home ranges in the study area. All plots were re-visited
at one-month intervals throughout the growing season (May to September). Forage
utilization was estimated by visual assessment of the amount of biomass removed by
herbivores – the ocular estimate-by-plot method (Pechanec and Pickford 1937; Irving et
al. 1995). This method for assessing biomass utilization is simple and the most reliable
under various species composition, dry matter and environmental circumstances (Tucker
1980), thus re-calibration is not necessary for each herbaceous community type. In
addition, at 40 of the 150 plots, forage utilization was also assessed using the pairedsubplot method (Bork and Werner 1999) in which vegetation biomass on an area exposed
72
to defoliation was compared with a nearby area whereon grazing had been excluded. At
these plots, I assessed the accuracy of my visual estimates of utilization by comparing
visual estimates to actual utilization (see Appendix E) in a linear regression. I found a
significant linear relationship and therefore concluded that visual estimates were useful
measures of forage utilization (Appendix E, Fig. E1).
4.2.6 Modeling Species Relationships in a Food Web
I overlaid all of my validated daytime and nighttime species distribution models
and the forage index in a GIS. At each 30-m pixel along roads and trails in the study area
(n=760,140) I sampled the human count, RSF values (wolves, elk and cattle) and forage
index. Furthermore, to investigate the top-down influence of wolf, elk and cattle
distribution on forage I sampled RSF values at plots where forage utilization was
measured (see above).
Relationships between species distribution models were tested using SEM,
specifically path analysis. Path analysis tests whether variables (i.e., species distributions)
in a hypothesized path are interrelated by examining the variances and covariances of the
variables (Kelloway 1998). Path analysis has previously been used to investigate species
relationships in food webs where the natural history of the system is sufficient to
construct a priori path diagrams (Elmhagen and Rushton 2007), which allows for
examination of directional processes involving complex and mediated relationships (i.e.,
ecosystem engineering versus trophic cascade hypotheses), as opposed to simply
bivariate cause-effect relationships (Kelloway 1998).
I used path analysis to investigate the influence of humans at multiple trophic
levels of the food web (forage, herbivores and carnivores) and test my two competing
73
hypotheses, whether humans influence species distributions as ecosystem engineers or by
causing trophic cascade. Humans were therefore modeled as exogenous variables
whereas other species were endogenous variables. Exogenous variables are starting
points of the model. Endogenous variables are determined by the pathways in the model,
and can serve as predictors for other endogenous variables. For example, elk distribution
could be predicted by forage distribution as well as predict wolf distribution. I developed
a SEM with pathways from humans to all trophic levels and compared SEMs with
ecosystem engineering pathways to SEMs trophic cascade pathways. Furthermore, I
compared these with simpler models with various human pathways removed and which
did not include pathways for which the association was non-significant.
I similarly conducted path analysis between wolf, elk and cattle RSF values with
forage utilization measured at the forage utilization plots (n=150). The influence of
wolves, elk and cattle on forage utilization were therefore tested in a separate SEM, but is
illustrated together with the SEM above. The forage utilization SEM could be considered
a sub-model of the full SEM to address specifically the question of wolf and herbivore
effects on forage utilization. In the forage utilization path analysis, wolves were treated as
exogenous variables and elk, cattle and forage utilization as endogenous.
I assessed SEM fit using the goodness-of-fit index (GFI), adjusted goodness-of-fit
index (AGFI), and root mean squared residual (RMR) and compared models using
Akaike Information Criterion (AIC). A GFI>0.9, AGFI>0.9 and with a similar value to
the GFI and RMR<0.05 indicate the model had a good fit to the data (Kelloway 1998).
The best pathway model was identified as that with the smallest AIC where all interaction
74
paths were significant and the model fit the data. Models were fitted using LISREL 8.72
(Joreskog and Sorbom 2005).
4.3 Results and Discussion
4.3.1 Humans Influence Elk and Cattle Distributions through Forage
My results indicated that herbivore species distributions (Fig. 4.1) in my
terrestrial ecosystem were driven by human influence on forage resources, and not by
trophic cascades mediated by wolves. The pathways in my parsimonious food web model
(Fig. 4.2) followed the ecosystem engineering hypothesis, as it was positive from human
counts to forage (β=0.637, p<0.0001, for day and night) and from forage to herbivores.
Elk distributions during the day (Fig. 4.1e) and night (Fig. 4.1f) were a positive function
of forage distribution (day: β=0.400, p<0.0001; night: β=0.369, p<0.0001; Fig. 4.2).
Cattle distribution was similarly related with forage (day: β=0.403, p<0.0001; night:
β=0.436, p<0.0001; Fig. 4.2) during the daytime (Fig. 4.1g) and nighttime (Fig. 4.1h). A
positive pathway from the predicted distributions of humans to cattle during the daytime
(β=0.220, p<0.0001; Fig. 4.2) and nighttime (β=0.175, p<0.0001; Fig. 4.2) may be an
indirect effect of the direct positive human influence on forage, and was consistent with
the ecosystem engineering hypothesis. Finally, I found positive pathways from herbivore
distribution to wolf distribution during the daytime (elk: β=0.293, p<0.0001, cattle:
β=0.303, p<0.0001; Fig. 4.2) and nighttime (elk: β=0.460, p<0.0001, cattle: β=0.482,
p<0.0001; Fig. 4.2). Similar pathways to those described above were found within two
wolf pack home ranges in the study area characterized as having the highest and lowest
average human counts on roads and trails within the home range boundaries (Appendix F,
Fig. F1).
75
(i.e., each class has approximately the same number of values and the change between intervals is consistent).
illustrative purposes roads and trails are exaggerated with a 1-km buffer and RSF values are binned into five geometric interval classes
habitat types and mean of the maximum Normalized Difference Vegetation Index within a 1 km radius of a 30-m2 pixel. For
measured from geographic information system datasets (Appendix C). The forage index is the product of the area of high-food-quality
models were calculated from resource selection functions (RSFs) using satellite- and GPS-telemetry data and habitat covariates
collected using road counters and trail cameras (Appendix B). Wolf (n = 14), elk (n = 62) and cattle (n = 50) day and night distribution
Alberta, Canada. Human distribution is calculated from an index of human density on roads and trails transformed into human counts
day (e) and night (f), cattle during the day (g) and night (h) and forage quality and quantity (i) along roads and trails in southwest
Figure 4.1. The spatial distribution of humans during the day (a) and night (b), wolves during the day (c) and night (d), elk during the
76
77
78
Figure 4.2. Structural equation model illustrating the direction and strength of
relationship between the spatial distribution of humans, wolves, elk, cattle and forage
quality and quantity and forage utilization during the daytime (sunrise to sunset) and
nighttime (sunset to sunrise)in southwest Alberta, Canada. Solid arrows indicate causal
direction of the consumer-resource interaction and line thickness is proportional to
relationship strength (path analysis β coefficient indicated). Human influences are
represented by dashed-dotted lines. Goodness of Fit Index (GFI), Adjusted Goodness of
Fit Index (AGFI), Root Mean Square Residual (RMR) and Akaike’s Information
Criterion (AIC) values of the models indicate model fit.
79
I did not measure the mechanism of how humans were acting as ecosystem
engineers. Obviously, human presence on roads and trails does not directly contribute to
forage. However, human population density drives the intensity of land use (Ellis and
Ramankutty 2008), thus human density is a useful proxy of human-caused land use
change. I suggest that in my study area, humans were positively contributing to forage in
the environment around them through range improvements (e.g., clearing forests and
enhancing productive grass species; Huston 1993), because livestock production is an
important historical and current economic activity (Alberta Beef Producers 2010).
Human counts (Fig. 4.1a and 4.1b) were generally higher in portions of the study area
where forage was the highest (Fig. 4.1i), suggesting that humans were driving forage
distribution. In fact humans are an important reason why parts of the world are green
(Polis 1999), because humans improve range productivity by inputting nutrients (e.g.,
fertilizer; Gu et al. 2009) and water into surrounding habitats.
4.3.2 Humans Negatively Influence Wolf Distribution, but this did not have Cascading
Effects
Humans had a direct effect on forage distribution that indirectly influenced other
species distributions in the food web (see above), but I identified a negative pathway
from humans to wolves (β=-0.228, p<0.0001 and β=-0.216, p<0.0001, for day and night,
respectively; Fig. 4.2), suggesting that humans also directly influenced wolf distribution.
Although not consistent with the ecosystem-engineering hypothesis, the negative
pathway from human to wolves was not surprising because of lethal control of wolves by
humans in southwest Alberta in response to livestock predation by wolves (Musiani et al.
2003). In areas where wolves are killed by humans, wolves will avoid humans across
80
space and time (Theuerkauf et al. 2003). Even in areas where they are protected from
humans (e.g., national parks) wolves avoid areas of high human density (Hebblewhite
and Merrill 2008).
I found similar evidence of spatiotemporal avoidance of humans by wolves in my
study. For example, within wolf home ranges (Appendix F, Fig. F1), the human to wolf
pathway was negative during the day (β=-0.102, p<0.0001) and absent during the night in
the high-human density wolf home range, and negative during the day (β=-0.097,
p<0.0001) but positive during the night (β=0.173, p<0.0001) in the low-human density
wolf home range. This suggested that when human density was low, there was no
negative effect of humans on wolf distribution. The spatiotemporal change in the
pathway from human to wolf distributions is also illustrated by the difference between
actual day and night locations of wolves monitored in this study (Fig. 4.3). During the
night, wolf locations tended to occur near roads (i.e., the darker pixels in Fig. 4.3), likely
due to better habitat for wolves there (Musiani et al. 2010) and the low use of roads by
humans over the night. Thus wolves in my study area likely avoided roads and trails with
high human counts to decrease their encounters with humans. Nevertheless, I did not find
strong evidence of a cascading top-down effect to herbivores from humans, mediated by
wolves. I identified a positive pathway from elk to wolf distributions and cattle to wolf
distributions during the day and night (see above), suggesting wolves were using areas
also used by prey, consistent with my ecosystem engineering hypothesis.
I found evidence of direct effects of cattle on forage utilization during the day
(β=0.178, p<0.05; Fig. 4.2) and night (β=0.169, p<0.05; Fig. 4.2) but I did not find
81
Figure 4.3. A sample case of wolf telemetry locations around roads with different human
counts during the daytime compared with nighttime in southwest Alberta, Canada.
Daytime locations are from sunrise to sunset, nighttime locations are sunset to sunrise.
Human count is the average number of humans on roads and trails during the day (a), and
night (b), calculated from an access index model (Appendix B). Darker pixels indicate
higher human counts.
82
evidence of direct effects on forage from elk. I also found a negative pathway from
wolves to forage utilization at night (β=-0.246, p<0.05; Fig. 4.2). Ultimately, it is likely
that the spatial distribution of forage utilization was not an indirect effect of humans on
herbivores that was mediated by wolves, but of humans directly managing cattle
distribution (pathway from humans to cattle, day: β=0.220, p<0.001, night: β=0.175,
p<0.001; Fig. 4.2). However, in areas and at times when humans were abundant (e.g.,
during the day), they might also exclude wolves from patches occupied by cattle (see
above), which is also consistent with human ecosystem engineering, as humans were
“engineering” predator-free space (a resource) for cattle by killing and displacing wolves.
The data used to create species distribution models did not completely overlap in
time (i.e., humans=2008, wolves=2004 to 2007, elk=2007 to 2009, cattle=2004 to 2006
and forage=2005). However, the distributions of each species did not change significantly
during the five year data collection period. For example, there were no significant road
developments or changes in human activity pattern, and telemetry data from wolves, elk
and cattle were typical of what was known about each species’ long-term distribution.
Furthermore, pooling data across years from several individual animals per year reduced
the potential influence of unusual distribution patterns of individual animals or years on
the averaged species distribution model. Finally, although forage biomass likely changes
from year-to-year, relative forage biomass across the study area was not expected to
change dramatically over five years (i.e., relatively high-forage areas remain that way
each year). We therefore concluded that distribution models were comparable.
83
4.3.3 Predominance of Bottom-up Effects in Human-Dominated Ecosystems
My results indicate that herbivore distribution was primarily influenced by forage.
In contrast, strong direct effects of predators on herbivores involving the same study
species were recently documented in other terrestrial ecosystems, specifically within the
North American national parks of Yellowstone (Ripple et al. 2001) and Banff
(Hebblewhite et al. 2005a). In my study area, range improvement by humans likely
contributed to stronger direct and indirect effects.
Previous research that addressed interactions between predator and vegetation
productivity forces in ecosystems found that predator effects might be weak or absent
when vegetation productivity is high, possibly due to increased potential for lower
trophic levels to compensate for predation (Schmitz 1994; Melis et al. 2009). Ultimately,
humans may tip the balance of predator versus forage regulation in ecosystems in favour
of forage by increasing vegetation productivity relative to predator biomass. Such human
impacts on food web structure might be common in agricultural areas where livestock
production and range improvement is a land-use objective.
Evidence of strong direct and indirect effects of predators on food webs in
protected areas has been used to argue for wolf reintroduction to many ecosystems
worldwide (Manning et al. 2009; Licht et al. 2010) including outside of large protected
areas, with the objective to restore those ecosystems. Human exclusion of predators has
likely influenced the function of many ecosystems by removing predation from
herbivores, which affects herbivory interactions. My results suggest that indirect effects
of humans on the food web through direct effects on forage can be greater than indirect
effects mediated by predators in human dominated landscapes. Thus I caution that
84
predator reintroduction with the objective to restore ecosystem structure and function
through predator regulation may not be effective in all ecosystems. Furthermore, in areas
where livestock production by humans is a dominant activity, humans may also influence
forage utilization indirectly by managing the density and distribution of domestic
herbivores. I suggest that all potentially strong direct and indirect human influences on
food webs be carefully considered for effective ecosystem conservation and management
85
CHAPTER FIVE: DIRECT AND INDIRECT EFFECTS OF HUMANS ON WHOLE
COMMUNITIES - HUMAN USE OF ROADS AND TRAILS HELPS PREY WIN THE
PREDATOR-PREY SPACE RACE
5.1 Introduction
In the previous chapters I documented the direct and indirect effects of humans on
a food web that included wolves, elk, cattle and vegetation. However, the food web in the
study area is more complex and includes several predator and prey species that interact
with each other. Humans might influence food webs through direct and indirect effects on
several species simultaneously. The notion that human presence might influence whole
multi-species assemblages has important implications for management and conservation
of entire communities and maintenance of their functionality.
The study aimed at determining whether humans influence the outcome of the
predator-prey “space race”. My purposes were to: (1) measure occurrence of large
mammals at camera traps placed along roads and trails; (2) test whether predators were
co-occurring with prey (i.e., predators maximizing spatial overlap), or prey were avoiding
predators (i.e., prey minimizing spatial overlap); and (3) test what factor(s), including
human presence, predators, food resources (i.e., forage for herbivore prey, prey counts for
predators) and habitat (i.e., forest cover and elevation) were influencing predator and
prey species occurrence. I used cluster and ordination analyses to describe species cooccurrence (including, wild large mammals, humans and cattle) and regression tree
analysis to test which factors influenced predator and prey species occurrence.
My rationale relies on using an appropriate means for assessing relative densities
of large mammals at different sites, which constitute pseudo-experimental replicates
86
receiving different human-use levels. Measuring the density and distribution of several
wide-ranging large mammalian species is challenging because it can be expensive,
invasive (if animals are captured, such as with telemetry studies) and labour intensive
(Mech and Barber 2002; MacKay et al. 2008). Recently, digital cameras traps have been
used as a relatively inexpensive and non-invasive means to measure large mammal
abundance and distribution (e.g., Kays and Slauson 2008; Rowcliffe et al. 2008; Rovero
and Marshall 2009). Furthermore, digital camera traps are indiscriminate, thus they can
provide information about several wildlife species in a region (e.g., Rowcliffe et al. 2008;
Rovero and Marshall 2009) and they can be used to measure the density and distribution
of humans too.
Comparing use of space by multiple species based on data from camera traps
requires some caution, as it can be biased towards larger (Tobler et al. 2008) and more
gregarious species (Treves et al. 2010) that might be more easily detected. Nevertheless,
in general the rate of photographing an animal at camera traps is correlated with animal
abundance and thus provides a useful index of species occurrence at a location (Carbone
et al. 2001; Carbone et al. 2002; but see Jennelle et al. 2002). Such an index should be
sufficient for comparing species co-occurrence at multiple camera sites collected during
the same period within the same study area.
Large terrestrial carnivores are generally sensitive to human disturbance
(Woodroofe 2000; Treves and Karanth 2003), therefore I hypothesized that predator
species would avoid high-human use roads and trails. I predicted that predators would not
occur at high density with humans at camera sites. Conversely, humans can provide
security cover from predation if prey are less sensitive to human disturbance than
87
predators (e.g., Berger 2007; Hebblewhite et al. 2005a; Beschta and Ripple 2009). In
addition, humans might provide food to herbivores through habitat enhancements (e.g., in
agricultural fields, around roads and trails Gordon 2009; Hegel et al. 2009). Therefore I
predicted that camera sites with high human use would have higher density of prey
species and lower density of predators.
5.2 Methods
5.2.1 Study Area
For a full description of the study area, please refer to chapter three, section 3.2.1.
5.2.2 Measuring Species Occurrence Using Camera Traps
To measure human, cattle and wildlife species occurrence, I deployed 43 digital
camera traps (RECONYX Silent ImageTM Model RM30, RECONYX© Inc., Holmen,
WI, USA). As the primary objective of the study was to document relative occurrence of
species where humans could also occur, areas around road and trails were ideal locations.
Cameras were placed within 5-meters of roads and trails ensuring that the view included
the area from the camera to the road/trail and an equal area at the other side.
It should be clearly noted here that the study did not rely on absolute measures of
density for any species. Obtaining species density information would require a different
methodological approach that accounts for habitat and sightability biases (e.g., Rowcliffe
et al. 2008). This study relies on relative indices of abundance independently gathered for
each study species among camera sites. Our methodological approach should allow
comparing counts of a given species at a given camera site to counts of the same species
at other camera sites. Unbiased sampling was achieved by producing 43 random points
within the study area using Hawth’s Tools (Beyer 2004) in ArcGIS 9.2 (ESRI© Inc.,
88
USA) and placing a camera along the nearest road or trail to each point. Cameras were
set to the highest sensor sensitivity with a delay of one picture per second and strapped to
trees using bungee cords and cable locks at a one meter height facing the trail/road.
Cameras were deployed from 17 April to 21 November 2008 for 7,421 trap days (mean
=173 trap days/camera). Thus, cameras measured the “summer” distribution of animals
only. Cameras were re-visited at one-month intervals to download data from memory
cards, change batteries and replace desiccant packs.
Detected species, and date and time of detection were recorded for each picture
taken. If multiple individuals were captured within a single photograph, each individual
was counted singularly. Multiple photographs within a short period of time (15 minutes)
that were obviously of the same animal were counted as one record, as suggested by other
camera trap studies (e.g., O’Brien et al. 2003). Indices of relative abundance (i.e.,
photographic rates) were calculated for each species to assess species occurrence at each
site. Abundance was indexed as the number of independent captures of each species per
100 trap-days (O’Brien et al. 2003; Balme et al. 2010).
5.2.3 Measuring Habitat at Camera Trap Sites
I used Geographic Information Systems (GIS) data to measure habitat
characteristics that might be important to large mammalian species occurrence, including
elevation and the amount of high-quality forage habitat and forest cover in the area
surrounding camera sites. From a digital elevation model (DEM) I calculated the average
elevation within a 1-km radius of each camera site.
To calculate the amount of high-quality forage habitat available to herbivores,
first I collapsed a 30-m2 spatial resolution 16-class vegetation cover GIS dataset
89
(McDermid et al. 2009) into two forage food-quality classes: high and low (Muhly et al.
2010b; chapter three) and calculated the area of high-food-quality forage habitat within a
1-km radius of each camera. Second, I obtained a 250-m2 spatial resolution dataset of the
maximum Normalized Difference Vegetation Index (NDVI) measured during the 2005
growing season. NDVI is an index of vegetation biomass (i.e., forage quantity) that is
useful to monitor the effect of vegetation on animals at large scales (Petorelli et al. 2005).
As a measure of forage quantity at each location, I calculated the mean of the maximum
NDVI values in a plant-growing season within a 1 km radius of each camera. I multiplied
this value by the area of high-quality forage habitat to obtain an index of forage quality
and quantity (hereafter, referred to as “forage”) within 1 km of each camera. To assess
the amount of security cover available to animals around camera sites, I further used the
vegetation cover GIS dataset to calculate the amount of actual forest within a 1-km radius
of each camera. Overall, classification accuracy of the vegetation map was 80%, as
calculated from ground-truthing of 245 independent, randomly selected test sites
surveyed in the field (McDermid et al. 2009).
5.2.4 Measuring Species Co-occurrence at Camera Sites
I tested whether species co-occurred at camera sites by conducting hierarchical
cluster analysis following McCune and Grace (2002). I defined each photographed
species as either present (i.e., detected) or absent (i.e., not detected) at each camera site
throughout the sampling period. I used a hierarchical agglomerative clustering strategy
and Ward’s linkage method to determine relatedness (-i.e. a statistical index of cooccurrence) among species presence, which are measure as Euclidean distances between
species. To obtain a graphical representation of species co-occurrence, a dendrogram was
90
produced with branches scaled with the percentage of information remaining in the
analysis (i.e., the longer the branch, the more species are separate).
I also tested for species co-occurrence at camera sites using nonmetric
multidimensional scaling (NMS) of species count data, a statistical approach that reduces
data into fewer dimensions (Young and Hamer 1987). NMS is the best choice for
ordination of data that does not meet the assumptions of multivariate normality, and it is
robust to large numbers of zero values (Minchin 1987; McCune and Grace 2002). Thus, it
is particularly appropriate for species counts, which may not be normally distributed and
can contain zeros for rare species. Also following McCune and Grace (2002), NMS
ordination was conducted using the following parameters: Sorenson distance measure,
500 iterations, optimum number of dimensions identified by a change in stress value <5,
and a Monte Carlo test run 250 times with randomized data. The software PCORD
version 5.17 (McCune and Mefford 2006) was used for both hierarchical cluster analysis
and NMS (see Mather 1976; Kruskal 1964 for algorithms used).
5.2.5 Influence of Humans, Predators, Food and Habitat on Species Occurrence
My study makes the reasonable assumption that predators might react to humans
differently than prey (e.g., Berger 2007; Hebblewhite et al. 2005a; Beschta and Ripple
2009). Therefore, I could aggregate species counts at camera sites into a predators’ guild
(i.e., wolves, cougars, grizzly bears and black bears) and prey guild (i.e., moose, elk,
white-tailed deer and mule deer). Coyotes were excluded from this analysis because they
are considered meso-carnivores, they rely on smaller prey and accordingly did not
associate strongly with the predator or prey guilds in this study (see Results, Fig. 5.2 and
5.3).
91
I used regression tree analysis (Breiman et al. 1984; De’ath and Fabricious 2000),
as a nonparametric approach to test whether humans, predators, food, and/or habitat
influenced predator and prey species counts at camera sites. Covariates considered in the
predator regression tree included humans (human counts at camera site), prey (sum of all
prey species), cattle, and habitat at the camera site (elevation, forage and forest cover,
measured in GIS, see above). Covariates considered in the prey regression tree included:
humans, predators (sum of all predator species), cattle and habitat at the camera site.
Regression trees recursively partition the dependent variable (i.e., predator or prey
count) into two comparatively homogeneous data clusters called nodes, and identify the
independent covariate (i.e., humans, predators, food or habitat) that best explains the
variation within each node. The optimum partition is determined by maximizing the
LogWorth statistic (i.e., the negative base 10 logarithm of the p-value calculated from the
sum of squares of the differences in means between the two groups formed by a partition;
Gaudard et al. 2006). Covariates in the regression tree can be re-used at each branch, thus
non-linear relationships may be identified. We used the regression tree as an exploratory
analysis; therefore we conducted recursive splitting of the tree to maximize significance
until a minimum of five terminal groups was reached (McCune and Grace 2002).
K-fold cross validation was used to assess regression tree model fit (Breiman et
al. 1984). The dataset was divided into 10 randomly assigned bins of data. Regression
trees were constructed using 9/10th of the dataset and the remaining bin was kept aside.
Predictions on species counts made by the regression tree were compared with data
observed in the remaining bin and the process was reiterated 10 times. Fit was
92
represented using R2 statistics. Regression tree analyses were conducted using JMP 7.0
software (SAS Institute©, Inc. 2007).
5.3 Results
5.3.1 Humans Co-occurred with Prey Species More than Predator Species
I obtained photographs (Fig. 5.1) of nine large mammalian wildlife species
including wolves, grizzly bears, cougars, black bears, coyotes, moose, elk, mule deer and
white-tailed deer. In Fig. 5.2, a dendrogram of the hierarchical cluster analysis of species
presence/absence data illustrates co-occurrence of species at camera sites. Percent
chaining of the cluster analysis was 23.08%, which is acceptable according to McCune
and Grace (2002). Predator species (i.e., wolves, grizzly bears, black bears and cougars)
were distinct from the wild prey/human group (0% of information and the longest
branches), indicating they did not typically co-occur at camera sites. Domestic cattle
(47% of information remaining) and coyotes (71% of information remaining) were more
closely clustered with wild prey/humans than predators. Humans were clustered with all
wild prey species (indicated by short branches), including from lowest to highest
association: elk and moose (89% of information remaining), mule deer (98% of
information remaining) and white-tailed deer (100% information remaining).
Species ordination scores, illustrated along axis one of Fig. 5.3, indicate the
relative co-occurrence of species at camera sites. The proportion of variance of the data
represented by axis one was 0.737. I obtained a stress value of 5.703 and instability of
<0.00001, which are both acceptable (McCune and Grace 2002). On axis one, two
species appear at the extremes: wolves (positive) and humans (negative). Other species of
large predators (grizzly bears, black bears, and cougars) were placed close to wolves,
93
Figure 5.1. A sample of photos taken by cameras deployed on roads and trails in
southwest Alberta, Canada during the summer of 2008. I photographed all large
mammalian species in southwest Alberta, also including: cougar (top left), wolf (top
right), moose (bottom left) and elk (bottom-right).
94
Figure 5.2. Dendrogram of the hierarchical cluster analysis of species presence/absence
data that illustrates co-occurrence of species at camera sites in southwest Alberta, Canada
during the summer of 2008. The dendrogram is scaled with the percentage of information
remaining in the analysis, where less information remaining indicates a weaker
association between species. Clusters were identified using the Ward’s linkage method
with the Euclidean distance measure.
95
Figure 5.3. Co-occurrence of species at camera sites as determined by non-metric
multidimensional scaling (NMS) ordination of species counts at camera sites in
southwest Alberta, Canada during the summer of 2008. Location along axis one where
the NMS score equals zero is indicated by a vertical dashed line. Ordinations along axis
one are indicated. Ordination was performed in PCORD version 5.17 using Sørenson’s
distance measure (McCune and Mefford 2006).
96
indicating potential co-occurrence of the predator guild. At the other extreme, humans
were most closely associated with domestic cattle. All prey species and coyotes were in
the middle, yet closer to predators than to humans and cattle.
5.3.2 Humans Co-Occurred with Prey but not with Predator Species
The regression tree model of prey (Fig. 5.4a) had a k-fold cross validation of
R2=0.370. In general, prey were three times more abundant on roads and trails with
>32.48 humans/day than on roads and trails with less humans. At the second level of the
tree, on roads and trails with less people, prey were twice as abundant in less forested
areas (i.e., where the percentage of forested area within 1 km of the cameras site was
36%) than forested areas. On the third level of the tree, in forested areas, prey were more
abundant on roads and trails with ≥0.03 predators/day than roads and trails with fewer
predators. Finally, on the fourth level of the tree, in forested areas with more predators,
prey were more abundant at lower elevations (<1,473 m), than higher elevations.
The regression tree model of predator count (Fig. 5.4b) had a k-fold cross
validation of R2=0.517. The first partition of the data indicated that predators were three
times more abundant on roads and trails with ≥0.26 prey/day than on roads and trails with
fewer prey. At the second level of the tree, on those roads and trails with more abundant
prey, predators were more abundant if there were ≥0.31 humans/day on roads and trails
compared with roads and trails with fewer humans. Also at the second level of the tree,
on roads and trails with <0.26 prey/day, predators were more abundant on roads and trails
with <1.44 humans/day than on roads and trails with more humans. However, at the third
level of the tree, predators were less abundant on roads trails if there were ≥18.71
humans/day. Finally, at the fourth level of the tree, predators were more abundant on
97
(a)
(b)
Figure 5.4. Regression tree analysis of large mammalian prey (top) and predator (bottom)
counts at camera sites in southwest Alberta, Canada during the summer of 2008. For each
of the nodes on the tree, the explanatory variable is shown with the value that best
determines the partition (i.e., the cut-off point that maximizes homogeneity within a
group). Also indicated for each node are the number of cameras in the group (count) and
the mean number of predator or prey photographs per 100 days (and standard deviation).
98
those roads and trails with <18.71 humans/day if there was ≥1.11 cattle/day, than on
roads and trails with fewer cattle.
5.4 Discussion
5.4.1 Spatial Separation of Predator and Prey Species
I did not find species from predator and prey guilds aggregated with each other in
the cluster dendrogram (Fig. 5.2) or ordination (Fig. 5.3). Rather, species within predator
and prey guilds tended to co-occur with species of the same guild at camera sites. These
results indicate partial spatial separation between predator and prey species and may
suggest that prey are “winning” the predator-prey “space race” (Sih 2005), i.e., prey
species were more effective at avoiding predators than predators were at tracking prey.
As a word of caution, predators may be selecting areas that improve their chance
at capturing prey rather than areas with high prey density (e.g., Hopcraft et al. 2005). In
addition, predators may adopt unpredictable patterns in their use of space to increase prey
uncertainty in perceiving predation risk of an area (e.g., Roth and Lima 2007). These
mechanisms, if present, limit the inference that prey are effectively avoiding being
predated by just displacing predators. However, the predator guild in this study
represented diverse hunting strategies. Typically, cougars are solitary ambush predators,
wolves are coursing predators that hunt in packs, and both black bears and grizzly bears
are solitary, omnivorous species that hunt opportunistically (Kays and Wilson 2009). It is
therefore unlikely that all predators in this study selected similar habitats that similarly
improve their chance at capturing prey, despite lower prey density there. Similarly
unpredictable patterns in all predators are not to be expected either. Therefore, I believe
99
that spatial separation of predators from prey likely resulted in prey effectively avoiding
predators.
5.4.2 Humans Tip the Predator-Prey Space Race in Favour of Prey
My results in this chapter indicate that at high densities, humans might displace
predators, providing a positive indirect effect on large mammalian herbivore species that
are less sensitive to humans. My results therefore support the hypothesis that humans can
help prey win the predator-prey space race by positively influencing prey species and
negatively influencing predator species distribution. Whereas prey were more abundant
on roads and trails with more humans (i.e., >32 people/day), predators were less abundant
on roads and trails that exceeded 18 humans/day, even if there were more prey there.
Furthermore, telemetry data collected from wolves (chapter four) and grizzly bears
(Northrup et al. In Review) in my study area confirm that predators avoid high-human
use areas. Other studies also suggest that human disturbance that displaces predator
species can benefit prey. For example, moose will use birthing areas close to high-human
use roads, as human disturbance deters grizzly bears from those areas (Berger 2007) and
ungulate density can be higher in areas with high-human density because carnivore
density is low there (e.g., Hebblewhite et al. 2005a; Beshcta and Ripple 2009). Such an
outcome is likely the result of large carnivore predators being more sensitive to human
disturbance than herbivores.
My results indicate species co-occurrence during the summer (i.e., April to
November) only. The relative density and distribution of each species may change
significantly during the winter. For example, predators actually might favour roads and
trails during the winter for ease of travel if roads and trails are ploughed free of snow or
100
snow is hard-packed by snow-machines (e.g., Kunkel and Pletscher 2001; Whittington et
al. 2005). It is also worth noting that memory can influence space use by mammals. In
particular, memory appears to play an important role in habitat selection and foraging by
ungulates (Bailey et al. 1996; Wolf et al. 2009). Memory of previous food, predator
interactions, security cover and human disturbance likely played a role in occurrence or
avoidance of certain camera sites by predator and prey species.
The positive association between herbivore prey and humans that I documented
might not only be the result of humans displacing predators, but also due to humans
improving forage around roads and trails (Trombulak and Frissell 2000). High quality
and quantity forage resources are correlated with high-human use roads and trails in the
study area (chapter four). Humans might therefore provide the best habitat patches for
herbivores by both deterring predators and improving food resources. Therefore, in my
study area large mammalian herbivores might be benefitting from security cover and food
resources provided by humans.
Studies that measure the combined effects of resources and predators on prey
species use of space are rare (but see Willems and Hill 2009), especially for large
mammalian species, because of the difficulty in collecting data on their distribution. My
study indicates that camera traps and GIS technologies are useful to simultaneously
document multiple species use of space and compare the relative influence of predators
and resources (e.g., forage) on prey species distribution. Furthermore, my study is unique
because in addition to resources and predators I considered the influence of humans on
species use of space.
101
CHAPTER SIX: CONCLUSIONS
6.1 The Economics of Livestock Predation by Wolves and Implications for Direct
Effects of Humans on Wolves
In chapter two my results indicate that lethal control of wolves in response to
livestock depredation is not motivated by economics (i.e., the costs of depredation to
producers). I suggest conducting further research to determine whether cultural, political
and/or ethical factors might be more significant at affecting tolerance for wolves (e.g.,
Montag 2003; Naughton-Treves et al. 2003). Until these broader issues are understood
and reconciled the negative direct effects of humans on wolves are likely to continue.
I found that the total number of cattle lost due to wolf depredation was minor
compared with other causes of death in Idaho, Montana and Wyoming. In addition,
livestock depredation costs, although increasing during the study period, were negligible
relative to income generated from livestock production in the study area. However, single
depredation events can potentially be costly to individual producers. Furthermore,
livestock producers in the area were facing significant economic challenges during this
period, as the price of cattle declined, the price of sheep remained stagnant, and the price
of both was highly variable. Given the unpredictability and overall decline in monetary
value of livestock, suspension of compensation programs may hamper those ranchers
affected by chronic wolf depredation and this might result in further opposition to wolf
presence. Thus, compensation programs should be continued in the future, as is common
practice for other carnivores that kill livestock worldwide (Wagner et al. 1997; Fourli
1999). However, compensation providers should be aware that such programs might be
limited in their ability for improving tolerance for wolves as surrogate issues related to
102
cultural, political and ethical factors might be more significant at affecting tolerance
(Montag 2003; Naughton-Treves et al. 2003).
In contrast to livestock prices, land prices were increasing and were relatively
stable from one year to the next. Thus, land value was much more predictable for
livestock producers. Demand for natural amenities (e.g., recreational opportunities,
viewscapes; sensu Hansen et al., 2002) has contributed to the trend in conversion of
agricultural land to higher density rural-residential land uses throughout the U.S.,
including the study area (Hansen et al. 2002; Sengupta and Osgood 2003; Brown et al.
2005; Gosnell and Travis 2005). The trends in land and livestock value I identified may
further encourage this recent phenomenon, as, from an economic perspective, selling land
would be the best way for livestock producers to profit from their assets. Wide scale
changes in land use, from agriculture to rural-residential, are possible given the current
economic conditions, which may have significant implications for wildlife conservation
and management. Such changes can dramatically alter habitat and increase human
presence in agricultural areas (Theobald et al. 1997; Hansen et al. 2002; Mitchell, et al.
2002) potentially resulting in a decline in ecosystem services (sensu Daily, 1997)
provided by private lands. Clearly the economic trends I identified and the literature
indicate this phenomenon deserves attention, particularly from groups interested in
conservation of species on agricultural lands.
Wolves are a public good, with a negative economic externality (livestock
depredation) shouldered by livestock producers. Ironically, large tracts of relatively
undeveloped land managed for livestock production may provide a positive externality
for wildlife conservation, including wolves; particularly relative to rural-residential
103
development. Therefore the ultimate question is how to reconcile these externalities.
Broader societal investment in preserving some forms of livestock production may be
necessary to avoid marginalization of rural communities and ensure ecosystem services,
including wolf conservation, contributed by some private lands continue to be provided
to the public.
6.2 Wolves Directly Affect Their Prey but Such Effects are Different for Domestic
Compared with Wild Animals
In chapter three I concluded that wolves have habitat-dependent risk effects on elk
and to a weaker extent on cattle. Both elk and cattle adjusted their use of high-foodquality habitat and security cover in response to wolves and therefore both species can
mediate the effect of predators on the food web where they co-occur. The types of risk
effects were different for each prey species, however, and the poor anti-predator response
of livestock likely reflected the consequences of domestication and artificial selection.
The ecological effects of predators such as wolves may be different in areas where
domestic animals such as cattle are the dominant herbivore in the ecosystem, if compared
with areas frequented by wild prey only. Mechanisms of predator-prey interactions and
the indirect effects on other species in ecosystems are well documented and considered in
conservation of wild animals and ecosystems (Hebblewhite et al. 2005a). However,
conservation groups as well as wildlife and rangeland managers should consider that such
effects might manifest differently in ecosystems where domestic animals are dominant
species.
How livestock respond to predator presence also has significant implications for
predator conservation in livestock production areas. Wolf predation on livestock causes a
104
significant human-wolf conflict, and the lethal control of wolves by humans is the typical
management response (see above). However, complete removal of wolves from cattle
range, or vice-versa, are unrealistic or unlikely without significant political repercussions
(Kellert et al. 1996). Thus, conflict between humans and wolves is likely to continue,
unless other means to break the cycle of depredation and lethal control of wolves are
identified. Various non-lethal approaches to livestock protection have been tried (see
Shivik 2006), but none has been universally applied. The attenuated anti-predator
response by cattle to wolves surely contributes to the cycle of livestock killed by wolves
and wolves killed by people as a consequence (see Musiani et al. 2003). Fluctuations in
densities of important predators such as wolves might have implications for risk effects
on prey and for the indirect effects on other species in ecosystems. My research indicates
similar considerations apply to scenarios where wolves are culled in response to livestock
depredation. Under these circumstances, people should be aware that changes in wolf
numbers might produce changes in the strength of ecosystem effects, and in the type of
effects if wild or domestic prey are dominant. My results on cattle selection for roads and
trails suggest livestock might perceive roads and trails as safe areas from predators
because they associate roads and trails with humans. A potential solution for protecting
livestock from wolves might therefore be to increase human presence on the landscape.
However, the efficacy of such a management action on livestock survival requires
experimental testing. Furthermore, increased human presence over large areas may
ultimately have negative effects on wolves and the ecosystem in general by excluding
wolves from those areas.
105
Another matter of consideration is that livestock producers are concerned about
the fitness consequences of the risk effects of predators in general on livestock, such as
increased stress and reduced foraging time (Howery and DeLiberto 2004). Further
research is required using quantitative techniques (e.g., Lind and Cresswell 2005) to
assess the energetic consequences of wolf visits to cattle: for example, whether risk
effects that I detected result in decreased weight gain. My results suggest that the
economic impacts of wolf-triggered risk effects on cattle could be considered while
planning compensation programs.
6.3 Direct and Indirect Effects of Humans Result in Ecosystem Engineering
Structuring Whole Food Webs
Ecosystem engineering by humans at multiple trophic levels of food webs might
be characteristic of most human-dominated ecosystems, especially compared with
protected areas that are generally less productive (Joppa and Pfaff 2009) and restrict
many types of human activities. Considering that only 13% (Jenkins and Joppa 2009) of
global lands are under protected status, my results confirm the dominant ecological role
of humans in shaping species interactions and food webs of terrestrial ecosystems.
Ecosystem-engineering by humans is likely to increase to meet the needs of a growing
human population. Subsequently, humans will have increasing impacts on ecosystem
structure and function through complex direct and indirect interactions with multiple
species in food webs of ecological communities. Documenting direct and indirect effects
of humans on several trophic levels of food webs will be necessary to understand fully
how humans influence ecosystem structure and function. Such information could be
crucial for conserving ecosystems as well as restoring them to desired states.
106
6.4 Direct and Indirect Effects of Humans can Also Influence Predator-Prey
Interactions
The outcome of a predator-prey “space race” is often influenced by a spatial
anchor, i.e., any environmental factor that is fixed in space that influences predator or
prey fitness (Sih 1984; Sih 2005). In chapter five, my results suggest that high-human use
roads and trails might be a positive spatial anchor to prey, providing refuge from
predators that are sensitive to human disturbance and potentially tipping the balance of
the predator-prey space race in favour of prey. Although my study hints at such a
mechanism, a greater understanding of the foraging strategies of each predator and prey
species is worth investigating to determine if humans are affecting also the ultimate
outcome of predator-prey interactions (i.e., predator and prey survival and fitness). In
addition, I cannot comment on whether human activity affects predator-prey interactions
off roads and trails. Nevertheless, roads and trails with high-human activity (i.e., >18
humans/day; Fig. 5.4) are pervasive throughout the study area (see chapter four) and
likely to increase, thus the effect of humans on predator-prey interactions has the
potential to be significant.
To effectively predict and mitigate the effects of human use of roads and trails on
food webs, managers must be aware of not only direct effects, but also potential indirect
effects on other species. Furthermore, lack of predators, and thus predator-prey
interactions, in an area can even have negative effects on biodiversity in general (Ray et
al. 2005). For example, direct and indirect effects of humans on predator and prey species
can indirectly influence some vegetation species via trophic cascades (e.g., Ripple et al.
107
2001) and also influence other species that utilize that vegetation, such as beavers and
song birds (Hebblewhite et al. 2005a; Ripple and Beshcta 2009).
6.5 Future Research Opportunities
Humans and their demand for resources are increasing in ecosystems. My results
clearly indicate that humans influence several species at multiple trophic levels of food
webs. Some influences are obvious, such as lethal control of wolves or range
improvements that increase vegetation biomass, but many influences are not so obvious,
for example animal domestication or the indirect effects of predator control and range
improvements on herbivores. My research is novel in that it attempts to quantify several
direct and indirect effects of humans on a food web, something resource managers and
scientists rarely do, despite acknowledging the importance of these effects. However,
while my research identifies some clear patterns in how humans may be influencing a
food web, there is a need for further work to identify the underlying mechanisms creating
those patterns. In fact, an important lesson from my work is that there is still a significant
need for research that quantifies the extent of human influences on food webs under a
variety of environmental conditions.
One example from my research that deserves greater research attention is
quantifying how humans influence herbivore species, specifically elk in my study area.
Ungulate conservation is a societal objective in many areas (Gordon et al. 2004). Human
impacts on ungulates are a concern in areas where demand for natural resources and
recreation is increasing, a common scenario in western North America (Galvin et al.
2008), including my study area (Arc Wildlife Services Ltd. 2004). To survive and
reproduce, ungulates must select habitat where they can obtain sufficient resources and at
108
the same time avoid predators (Festa-Bianchet 1988; Houston et al. 1993). However,
humans also influence ungulate habitat selection and they can do so directly, through
hunting and displacement (Stankowich 2008; Fryxell et al. 2010), and indirectly, by
improving resources (for example, through agriculture – see Chapter 4) or reducing
predation (by killing and displacing ungulate predators – see Chapter 5). Although these
different direct and indirect effects of humans are conceptually understood (Peek 1980),
very little research has attempted to disentangle and quantify the influence of direct and
indirect effects of humans on ungulate habitat selection. Yet understanding such effects
would improve ungulate conservation in increasingly human dominated landscapes.
Future research could test some hypotheses on how different direct and indirect
effects of humans influence herbivore habitat selection. For example, in southwest
Alberta, elk selection of food resources, predation risk, and human density could be
tested in three distinct land-use areas (national parks, provincial public land and private
land) where different direct and indirect effects of humans occur. In national parks there
is no hunting (direct effects), agriculture (resource-mediated effects), or predator-control
(predator-mediated effects), thus humans might have limited influence on elk, as elk
select optimal (high-resource and low-predation risk) habitat regardless of human
density. Alternatively, on provincial public lands, where agriculture is limited, predatorcontrol is occasional and hunting pressure on elk is high, humans might have strong
direct effects on elk. Thus, in areas and periods of low-human density, elk might select
optimal habitat. However, in areas and periods of high-human density, elk might select
sub-optimal (low-resource and high-predation risk) habitat, due to trade-offs being made
between human density and resources and human density and predation risk. Finally, on
109
private lands where agriculture and predator-control are extensive but hunting pressure
on elk is low, humans might have strong indirect effects on elk. Elk might therefore
select high-resource habitat and disregard predation risk (i.e., no avoidance) regardless of
human density, as humans provide food and predator-free space to elk.
Wildlife species must make complex decisions regarding resource acquisition,
predator avoidance and several human influences when selecting habitat. Research that
disentangles the effects of humans could demonstrate for the first time how wildlife not
only select and trade-off between resources and predation risk, but also different types of
human influences. Such results could demonstrate the mechanisms of direct and indirect
effects of humans on wildlife and help managers better predict the impact of different
types of land-use activities on wildlife populations.
In this thesis, I documented direct and indirect effects of humans on several
species that interact in a terrestrial food web. To fully understand and where necessary
mitigate the impact of humans on species in an ecosystem, managers must not limit their
considerations to direct effects, but also consider the indirect effects on species at several
trophic levels of food webs.
110
REFERENCES
Abrams, P.A. 1995. Implications of dynamically variable traits for identifying,
classifying, and measuring direct and indirect effects in ecological communities. The
American Naturalist 146:112-134.
Agrawal, A.A. 2001. Phenotypic plasticity in the interactions and evolution of species.
Science 294:321-326.
Albers, H. and Ferraro, P. 2006. Economics of biodiversity conservation in developing
countries. In: M. Toman and R. Lopez (Editors), Economic development and
environmental sustainability: new policy options. Oxford University Press, New
York, pp. 382–411.
Alberta Beef Producers 2010. Alberta Cattle Industry Statistics. Available:
http://albertabeef.org/industry/industry-information/. Accessed 28 April 2010.
Alberta Sustainable Resource Development (ASRD) 2002. Southern foothill-mountain
elk survey. Alberta Sustainable Resource Development, Fish and Wildlife Division,
Blairmore, Alberta, pp. 38.
Alberta Sustainable Resource Development (ASRD) and Alberta Conservation
Association (ACA) 2010. Status of the grizzly bear (Ursus arctos) in Alberta: Update
2010. ASRD Wildlife Status Report 37:1-44.
Alcock, J. 2005. Animal Behaviour. Sinauer Associates, Inc., pp. 546
Apps, C.D., Mclellan, B.N., Woods, J.G. and Proctor, M.F. 2004. Estimating grizzly bear
distribution and abundance relative to habitat and human influence. Journal of
Wildlife Management 68:138–152.
111
Arc Wildlife Services Ltd. 2004. Selected ecological resources of Alberta’s Castle
Carbondale: A synopsis of current knowledge. Report prepared for CPAWS
Calgary/Banff and Shell Canada. 216 pp.
Bailey, D.W., Gross, J.E., Laca, E.A., Rittenhouse, L.R., Coughenour, M.B., Swift, D.M.
and Sims P.L. 1996. Mechanisms that result in large herbivore grazing distribution
patterns. Journal of Range Management 49:386-400.
Ballard, W.B., Whitman, J.S. and. Gardner, C.L 1987. Ecology of an exploited wolf
population in south-central Alaska. Wildlife Monographs 98:1-54.
Balme, G.A., Slotow, R. and Hunter, L.T.B. 2010. Edge effects and the impact of nonprotected areas in carnivore conservation: leopards in the Phinda–Mkhuze Complex,
South Africa. Animal Conservation 13:315–323.
Bangs, E.E., Fontaine, J. A., Smith, D. W., Murphey, K.M., Mack, C. M. and Niemeyer,
C.C. 1998. Status of gray wolf restoration in Montana, Idaho, and Wyoming. Wildlife
Society Bulletin 26:785-798.
Bangs, E.E., Fontaine, J.A., Jimenez, M.D., Meier, T.J., Bradley, E.H., Niemeyer, C.C.,
Smith, D.W., Mack, C.M., Asher, V. and Oakleaf, J.K. 2004. Managing wolf/human
conflict in the northwestern United States. In: R. Woodroffe, S. Thirgood and A.
Rabinowitz (Editors), People and Wildlife: Coexistence or Conflict? Cambridge
University Press, Cambridge, United Kingdom, pp. 340-356.
Bangs, E.E. and Shivik, J. 2001. Managing wolf conflict with livestock in the
Northwestern United States. Carnivore Damage Prevention News 3:2-5.
Barnosky, A.D., Koch, P.L., Feranec, R.S., Wing, S.L. and Shabe, A.B. 2004. Assessing
the causes of late Pleistocene extinctions on the continents. Science 306:70-75.
112
Bastian, C.T., McLeod, D.M., Germino, M.J., Reiners, W.A. and Blasko, B.J. 2002.
Environmental amenities and agricultural land values: a hedonic model using
geographic information systems data. Ecological Economics 40:337–349.
Baumol, W.J. 1972. On taxation and the control of externalities. The American Economic
Review 62:307-322.
Beaulieu, M.S. and Bédard, F. 2003. A geographic profile of Canadian livestock, 19912001. Agriculture and Rural Working Paper Series Working Paper No. 62, Statistics
Canada, Agriculture Division.
Beaulieu, M.S., Bédard, F. and Lanciault. P. 2001. Distribution and concentration of
Canadian livestock. Agriculture and Rural Working Paper Series Working Paper No.
47, Statistics Canada, Agriculture Division.
Beck, J.L. and Peek, J.M. 2005. Diet composition, forage selection, and potential for
forage competition among elk, deer, and livestock on aspen–sagebrush summer
range. Rangeland Ecology and Management 58:135–147.
Beckerman, A.P., Uriarte, M. and Schmitz, O.J.. 1997. Experimental evidence for a
behaviour-mediated trophic cascade in a terrestrial food chain. Proceedings of the
National Academy of Sciences 94:10735–10738.
Bedunah, D. and Willard, E. E. 1987. Importance of forest lands to ranching in Montana.
Rangelands 9:168-170.
Berg, R.T. and Butterfield R.M. 1976. New concepts of cattle growth. Sydney University
Press, Sydney, Australia, pp. 240.
Berger, J. 2007. Fear, human shields and the redistribution of prey and predators in
protected areas. Biology Letters 3:620–623.
113
Berger, K.M. 2006. Carnivore-livestock conflicts: effects of subsidized predator control
and economic correlates on the sheep industry. Conservation Biology 20:751–761.
Beschta, R.L. and Ripple, W.J. 2009. Large predators and trophic cascades in terrestrial
ecosystems of the western United States. Biological Conservation 142:2401–2414.
Beyer, H.L. 2004. Hawth's Analysis Tools for ArcGIS. Available:
http://www.spatialecology.com/htools. Accessed 28 April 2010.
Boitani, L. 1982. Wolf management in intensively used areas of Italy. In F.H. Harrington
and P.C. Paquet (Editors), Wolves of the world: perspectives of behaviour, ecology,
and conservation. Noyes Publications, Park Ridge, New Jersey, pp. 158-172.
Bolker, B.M., Brooks, M.E., Clark, C.J. Geange, S.W., Poulsen, J.R., Stevens, M.H.H.
and White, J.-S.S. 2009. Generalized linear mixed models: a practical guide for
ecology and evolution. Trends in Ecology and Evolution 24: 127-135.
Bork, E.W. and Werner S.J. 1999. Implications of spatial variability for estimating forage
use. Journal of Rangeland Management 52:151–156.
Bostedt, G. and Grahn, P. 2008. Estimating cost functions for the four large carnivores in
Sweden. Ecological Economics 68:517–524.
Boyce, M.S., Vernier, P.R., Nielsen, S.E. and Schmiegelow, F.K.A. 2002 Evaluating
resource selection functions. Ecological Modelling 157:281-300.
Breiman, L., Friedman, J.H., Olshen, R.A. and Stone, C.J. 1984. Classification and
regression trees. Wadsworth International, Belmont, CA, pp. 358.
Brown, D.G., Johnson, K.M., Loveland, T.R. and Theobald, D.M. 2005. Rural land-use
trends in the conterminous United States, 1950–2000. Ecological Applications
15:1851–1863.
114
Brown, J.S. 1999. Vigilance, patch use, and habitat selection: foraging under predation
risk. Evolutionary Ecology Research 1:49–71.
Butterfield, R.M. 1988. New concepts of sheep growth. University of Sydney Press,
Sydney, Australia, pp. 168.
Carbone, C., Christie, S., Conforti, K., Coulson, T., Franklin, N., Ginsberg, J.R.,
Griffiths, M., Holden, J., Kawanishi, K., Kinnaird, M., Laidlaw, R., Lynam, A.,
Macdonald, D.W., Martyr, D., McDougal, C., Nath, L., O'Brien, T., Seidensticker, J.,
Smith, D.J.L., Sunquist, M., Tilson, R. and Wan Shahruddin, W.N. 2001. The use of
photographic rates to estimate densities of tigers and other cryptic mammals. Animal
Conservation 4:75-79.
Carbone, C., Christie, S., Conforti, K., Coulson, T., Franklin, N., Ginsberg, J.R.,
Griffiths, M., Holden, J., Kawanishi, K., Kinnaird, M., Laidlaw, R., Lynam, A.,
Macdonald, D.W., Martyr, D., McDougal, C., Nath, L., O'Brien, T., Seidensticker, J.,
Smith, D.J.L., Sunquist, M., Tilson, R. and Wan Shahruddin, W.N. 2002. The use of
photographic rates to estimate densities of cryptic mammals: response to Jennelle et al.
Animal Conservation 5:121–123.
Carbyn, L.N. 1983. Wolf predation on elk in Riding Mountain National Park, Manitoba.
Journal of Wildlife Management 47:963-976.
Carroll, C., Phillips, M.K., Schumaker, N.H. and Smith, D.W. 2003. Impacts of
landscape change on wolf restoration success: Planning a reintroduction program
based on static and dynamic spatial models. Conservation Biology 17:536-548.
115
Chambers, C.M. and Whitehead, J.C. 2003. A contingent valuation estimate of the
benefits of wolves in Minnesota. Environmental and Resource Economics 26:249–
267.
Chapin III, F.S., Walker, B.H., Hobbs, R.J., Hooper, D.U., Lawton, J.H., Sala, O.E. and
Tilman, D. 1997. Biotic control over the functioning of ecosystems. Science 277:500504.
Christianson, D. and Creel, S. 2010 A nutritionally mediated risk effect of wolves on elk.
Ecology 91:1184–1191.
Ciucci, P. and Boitani, L. 1998. Wolf and dog depredation on livestock in central Italy.
Wildlife Society Bulletin 26:504-514.
Collinge, M. 2008. Relative risks of predation on livestock posed by individual wolves,
black bears, mountain lions, and coyotes in Idaho. In: R.M. Timm and M.B. Madon,
(Editors), Proceedings of the 23rd Vertebrate Pest Conference. University of
California, Davis, pp. 129-133.
Creel, S. and Christianson, D. 2008. Relationships between direct predation and risk
effects. Trends in Ecology and Evolution 23:194-201.
Creel, S., Winnie Jr., J., Maxwell, B., Hamlin, K. and Creel M. 2005. Elk alter habitat
selection as an antipredator response to wolves. Ecology 86:3387–3397.
Crête, M. and Manseau, M. 1996. Natural regulation of cervidae along a 1000 km
latitudinal gradient, change in trophic dominance. Evolutionary Ecology 10:51-62.
Daily, G.C. 1997. Nature’s services: societal dependence on natural ecosystems. Island
Press, Washington, D.C., pp. 392.
116
Dambrine, E., Dupouey, J-L., Laut, L., Humbert, L., Thinon, M., Beaufils, T. and
Richard, H. 2007. Present forest biodiversity patterns in France related to former
Roman agriculture. Ecology 88:1430–1439.
De’ath, G. and Fabricius, K.E. 2000. Classification and regression trees: a powerful yet
simple technique for ecological data analysis. Ecology 81:3178–3192.
DelGiudice, G.D. 1998. Surplus killing of white-tailed deer by wolves in northcentral
Minnesota. Journal of Mammalogy 79:227-235.
Diamond, J. 2002. Evolution, consequences and future of plant and animal domestication.
Nature 418:700-707.
Duffield, J. 1992. An economic analysis of wolf recovery in Yellowstone: Park visitor
attitudes and values. In: J.D. Varley and W.G. Brewster (Editors), Wolves for
Yellowstone? A report to the United States Congress, Vol. 4, Research and analysis.
U.S. Department of the Interior, National Park Service, Washington, D.C., pp. 235287.
Durbin, T.D., Smith, M.R., Wilson, R.D. and Rhee, S.H. 2004. In-use activity
measurements for off-road motorcycles and all-terrain vehicles. Transportation
Research Part D 9:209–219.
Eibl-Eibesfeld, I. 1975. Ethology: biology of behaviour. Holt, Rinehart and Winston,
New York, NY, pp. 544.
Ellis, E.C. and Ramankutty, N. 2008. Putting people in the map: anthropogenic biomes of
the world. Frontiers in Ecology and Environment 6:439–447.
Elmhagen, B. and Rushton, S.P. 2007. Trophic control of mesopredators in terrestrial
ecosystems: top-down or bottom-up? Ecology Letters 10:197-206.
117
ESRI (2009) ArcGIS 9.2. Redlands, California.
Estes, J.A. 1996. Predators and ecosystem management. Wildlife Society Bulletin
24:390-396.
Festa-Bianchet, M. 1988. Seasonal range selection in bighorn sheep: conflicts between
forage quality, forage quantity, and predator avoidance. Oecologia 75:580-586.
Fico, R., Morsetti, G. and Giovannini, A. 1993. The impact of predators on livestock in
the Abruzzo region of Italy. Revue Scientifique et Technique de l'Office International
Des Epizooties 12:39-50.
Fieberg, J., Matthiopoulos, J., Hebblewhite, M., Boyce, M.S. and Frair, J.L. 2010.
Autocorrelation and studies of habitat selection: problem, red herring, or opportunity?
Philosophical Transactions of Royal Society B-Biological Sciences 365:2233–2244
Flaxman, S.M. and Lou, Y. 2009. Tracking prey or tracking the prey’s resource?
Mechanisms of movement and optimal habitat selection by predators. Journal of
Theoretical Biology 256:187–200.
Foley, J.A., DeFries, R., Asner, G.P., Barford, C., Bonan, G, Carpenter, S.R., Chapin,
F.S., Coe, M.T., Daily, G.C. Gibbs, H.K., Helkowski, J.H., Holloway, T., Howard,
E.A. Kucharik, C.J. Monfreda, C., Patz, J.A., Prentice, I.C., Ramankutty, N. and
Snyder, P.K. 2005. Global consequences of land use. Science 309: 570-574.
Foley, C.W., Seerly, R.W., Hansen, W.J. and Curtis, W.E. 1971. Thermoregulatory
response to cold environment by neonatal wild and domestic piglets. Journal of
Animal Science 32:926–929.
118
Fortin, D., Beyer, H.L. Boyce, M.S., Smith, D.W., Duchesne, T. and Mao, J.S. 2005.
Wolves influence elk movements: behaviour shapes a trophic cascade in Yellowstone
National Park. Ecology 86:1320–1330.
Fourli, M. 1999. Compensation for damage caused by bears and wolves in the European
Union. DG XI, European Commission, Bruxelles, Belgium.
Frair, J.L., Merrill, E.H., Visscher, D.R., Fortin, D., Beyer, H.L. and Morales, J.M. 2005.
Scales of movement by elk (Cervus elaphus) in response to heterogeneity in forage
resources and predation risk. Landscape Ecology 20:273–287.
Frid, A. and Dill L.M. 2002. Human-caused disturbance stimuli as a form of predation
risk. Conservation Ecology 6:11, http://www.consecol.org/vol6/iss1/art11/
Fritts, S.H., Bangs, E.E. and Goreb, J.F. 1994. The relationship of wolf recovery to
habitat conservation and biodiversity in the northwestern United States. Landscape
and Urban Planning 28:23-32.
Fritts, S.H., Paul, W.J. Mech, L.D. and Scott, D.P. 1992. Trends and management of
wolf-livestock conflicts in Minnesota. U.S. Fish and Wildlife Service Resource
Publication 181:1-27.
Fritts, S.H., Stephenson, R.O., Hayes, R.D. and Boitani, L. 2003. Wolves and humans. In:
L.D. Mech and L. Boitani (Editors), Wolves: behaviour, ecology, and conservation.
University of Chicago Press, Chicago, IL, pp. 289-316.
Fryxell, J.M., Packer, C., McCann, K., Solberg, E.J. and Saether B.-E. 2010. Resource
management cycles and the sustainability of harvested wildlife populations. Science
328:903-905.
119
Fuller, T.K. 1989. Population dynamics of wolves in north-central Minnesota. Wildlife
Monographs 105:1-41.
Galvin, K.A. Reid, R.S. Behnke Jr., R.H. and Hobbs, N.T. 2008. Fragmentation in semiarid and arid landscapes: consequences for human and natural systems. Springer, The
Netherlands. 411 pp.
Gaudard, M., Ramsey, P. and Stephens, M. (2006) Interactive data mining and design of
experiments: The JMP partition and custom design platforms. North Haven Group,
LLC: Cary, N.C.
Gillies, C., Hebblewhite, M., Nielsen, S.E., Krawchuk, M.A., Aldridge, C.L., Frair, J.L.,
Saher, D.J., Stevens, C.E. and Jerde, C.L. 2006. Application of random effects to the
study of resource selection by animals. Journal of Animal Ecology 75:887–898.
Gipson, P.S., Ballard, W.B. and Nowak, R.M. 1998. Famous North American wolves and
the credibility of early wildlife literature. Wildlife Society Bulletin 26:808-816.
Gipson, P.S. and W.J. Paul. 1994. Wolves: prevention and control of wildlife damage.
University of Nebraska Press, Lincoln, NE, pp. 129.
Godvik, I.M.R., Loe, L.E., Vik, J.O., Veiberg,V., Langvatn, R. and Mysterud, A. 2009.
Temporal scales, trade-offs, and functional responses in red deer habitat selection.
Ecology 90:699–710.
Gordon, I.J., Hester, A.J., and Festa-Bianchet, M. 2004. The management of wild large
herbivores to meet economic, conservation and environmental objectives. Journal of
Applied Ecology 41:1021-1031.
Gordon, I.J. 2009. What is the future for wild, large herbivores in human-modified
agricultural landscapes? Wildlife Biology 15:1-9.
120
Gosnell, H. and Travis, W.R. 2005. Ranchland ownership dynamics in the Rocky
Mountain West. Rangeland Ecology and Management 58:191–198.
Grandin, T. and Deesing, M.J. 1998. Genetics and behaviour during handling, restraint,
and herding. In: T. Grandin (Editor), Genetics and behaviour of domestic animals.
Academic Press, New York, NY, pp. 113-144
Gross, M.R. 1998. One species with two biologies: Atlantic salmon (Salmo salar) in the
wild and in aquaculture. Canadian Journal of Fisheries and Aquatic Science 55:131–
144.
Gu, B., Chang, J., Ge, Y., Ge, H., Yuan, C., Peng, C. and Jiang, H. 2009. Anthropogenic
modification of the nitrogen cycling within the Greater Hangzhou Area system,
China. Ecological Applications 19:974-988.
Gude, J.A. Garrott, R.A., Borkowski, J.J. and King, F. 2006. Prey risk allocation in a
grazing ecosystem. Ecological Applications 16:285–298.
Gunson, J.R. 1983. Status and management of wolves in Alberta. In L.N. Carbyn
(Editor), Wolves in Canada and Alaska. Canadian Wildlife Service Report 45,
Canadian Wildlife Service, Ottawa, ON, pp. 25-29
Gusset, M., Swarner, M.J., Mponwane, L., Keletile, K. and McNutt, J.W. 2009. Human–
wildlife conflict in northern Botswana: livestock predation by endangered African
wild dog Lycaon pictus and other carnivores. Oryx 43:67-72.
Halliday, T.R. 1980. The extinction of the passenger pigeon Ectopistes migratorius and
its relevance to contemporary conservation. Biological Conservation 17:157-162
Hansen, A.J., Rasker, R., Maxwell, B., Rotella, J.J. Johnson, J.D., Parmenter, A.W.,
Langner, U., Cohen, W.B., Lawrence, R.L. and Kraska, M.P.V. 2002. Ecological
121
causes and consequences of demographic change in the new west. BioScience
52:151-162.
Hanson, J.D., Hendrickson, J. and Archer, D. 2008. Challenges for maintaining
sustainable agricultural systems in the United States. Renewable Agriculture and
Food Systems, Published online by Cambridge University Press
doi:10.1017/S1742170507001974.
Hayes, R.D., Baer, A. and Larsen, D.L. 1991. Population dynamics and prey relationships
of an exploited and recovering wolf population in the southern Yukon. Yukon Fish
and Wildlife Branch Final Report TR-91-1, Whitehorse, YK.
Hayes, R.D., Baer, A.M., Wotschikowsky, U. and Harestad, A.S. 2000. Kill rate by
wolves on moose in the Yukon. Canadian Journal of Zoology 78:49-59.
Hebblewhite, M., White, C.A., Nietvelt, C.G., Mckenzie, J.A., Hurd, T.E., Fryxell, J.M.,
Bayley, S.E and Paquet, P.C. 2005a. Human activity mediates a trophic cascade
caused by wolves. Ecology 86: 2135–2144.
Hebblewhite, M., Merrill, E. H. and McDonald, T.L. 2005b. Spatial decomposition of
predation risk using resource selection functions: an example in a wolf-elk predatorprey system. Oikos 111:101-111.
Hebblewhite, M., Merrill, E. H., Morgantini, L.E., White, C.A., Allen, J.R., Bruns, E.,
Thurston, L. and Hurd, T.E. 2006. Is the migratory behaviour of montane elk herds in
peril? The case of Alberta’s Ya Ha Tinda elk herd. Wildlife Society Bulletin 34:12801294.
122
Hebblewhite, M. and Merrill, E.H. 2008. Modelling wildlife–human relationships for
social species with mixed-effects resource selection models. Journal of Applied
Ecology 45:834–844.
Hegel, T.M., Gates, C.C. and Eslinger, D. 2009. The geography of conflict between elk
and agricultural values in the Cypress Hills, Canada. Journal of Environmental
Management 90:222-235.
Heithaus, M.R., Wirsing, A.J., Burkholder, D., Thomson, J. and Dill, L.M. 2009.
Towards a predictive framework for predator risk effects: the interaction of landscape
features and prey escape tactics. Journal of Animal Ecology 78:556-562.
Hell, P. 1993. Current situation and perspectives of the wolf in Czechoslovakia. In: C.
Promberger, and W. Schroeder (Editors), Wolves in Europe: status and perspectives.
Munich Wildlife Society, Munich, Germany, pp. 36-42.
Hobbs, N.T. 1996. Modification of ecosystems by ungulates. Journal of Wildlife
Management 60:695-713.
Hobbs, N.T., Galvin, K.A., Stokes, C.J., Lackett, J.M., Ash, A.J., Boone, R.B., Reid, R.S.
and Thornton, P.K. 2008. Fragmentation of rangelands: Implications for humans,
animals, and landscapes. Global Environmental Change 18:776–785.
Holling, C.S. 1965. The functional response of predators to prey density and its role in
mimicry and population regulation. Memoirs of the Entomological Society of Canada
45:3-60.
Hopcraft, J.G.C., Sinclair, A.R.E. and Packer , C. 2005. Planning for success: Serengeti
lions seek prey accessibility rather than abundance. Journal of Animal Ecology 74:559–
566.
123
Houston, A.I., McNamara, J.M., and Hutchinson, J.M.C. 1993. General results
concerning the trade-off between gaining energy and avoiding predation.
Philosophical Transactions: Biological Sciences 341:375-397.
Howery, L.D. and DeLiberto, T.J. 2004. Indirect effects of carnivores on livestock
foraging behaviour and production. Sheep and Goat Research Journal 19:53-57.
Huggard, D.J. 1993. Prey selectivity of wolves in Banff National Park: I. Prey species.
Canadian Journal of Zoology 71:130-139.
Hunter, L.T.B. and Skinner, J.D. 1997. Vigilance behaviour in African ungulates: the role
of predation pressure. Behaviour 135:195-211.
Hunter, M.D. and Price, P.W. 1992. Playing chutes and ladders: heterogeneity and the
relative roles of bottom-up and top-down forces in natural communities. Ecology
73:724-732.
Husseman, J.S., Murray, D.L., Power, G., Mack, C., Wenger, C.R. and Quigley, H. 2003.
Assessing differential prey selection patterns between two sympatric large carnivores.
Oikos 101:591–601.
Huston, M. 1993. Biological diversity, soils, and economics. Science 262:1676-1680.
Irving, B.D., Rutledge, P.L., Bailey, A.W., Naeth, M.A. and Chanasyk, D.S. 1995. Grass
utilization and grazing distribution within intensively managed fields in Central
Alberta. Journal of Range Management 48:358-361.
Jenkins, C. and Joppa, L. 2009. Expansion of the global terrestrial protected area system.
Biological Conservation 142:2166–2174.
124
Jennelle, C.S., Runge, M.C. and MacKenzie, D.I. 2002. The use of photographic rates to
estimate densities of tigers and other cryptic mammals: a comment on misleading
conclusions. Animal Conservation 5:119–120.
Johnson, C.J., Nielsen, S.E., Merrill, E.H., McDonald, T.L. and Boyce, M.S. 2006.
Resource selection functions based on use-availability data: theoretical motivation
and evaluation method. Journal of Wildlife Management 70:347–357.
Johnson, C.N. 2010. Red in tooth and claw: how top predators shape terrestrial
ecosystems. Journal of Animal Ecology 79:723–725.
Johnson, D.H. 1980. The comparison of usage and availability measurements for
evaluating resource preference. Ecology 61:65-71.
Johnsson, J.I., Höjesjö, J. and Fleming, I.A. 2001. Behavioural and heart rate responses to
predation risk in wild and domesticated Atlantic salmon. Canadian Journal of
Fisheries and Aquatic Science 58:788–794.
Jones, C.G., Lawton, J.H. and Shachak, M. 1997. Positive and negative effects of
organisms as physical ecosystem engineers. Ecology 78: 1946-1957.
Joppa, L.N. and Pfaff, A. 2009. High and far: Biases in the location of protected areas.
PLoS ONE 4: e8273. doi:10.1371/journal.pone.0008273.
Joreskog, K.G. and Sorbom, D. 2005. LISREL 8.72 .Lincolnwood, IL Scientific Software
International, Inc.
Kays, R.W. and Slauson K.M. 2008. Remote cameras. In: R.A. Long, P. MacKay, W.J.
Zielinski and J.C. Ray (Editors), Noninvasive survey methods for carnivores Island
Press, Washington, D.C., pp. 110-140.
125
Kays, R.W. and Wilson, D.E. 2009. Mammals of North America. Princeton University,
Princeton, NJ, pp. 240.
Keeling, L. and Gonyou, H. 2001. Social behaviour in farm animals. CAB International
Publishing, Oxon, UK, pp. 413.
Kellert, S.R., Black, M., Rush, C.R. and Bath, A.J. 1996. Human culture and large
carnivore conservation in North America. Conservation Biology 10:977-990.
Kelloway, E.K 1998. Using LISREL for structural equation modeling. Thousand Oaks,
California, Sage Publications, 147 p.
Kluever, B.M., Howerya, L.D., Breck, S.W. and Bergman, D.L. 2009. Predator and
heterospecific stimuli alter behaviour in cattle. Behavioural Processes 81:85–91.
Koper, N. and Manseau, M. 2009. Generalized estimating equations and generalized
linear mixed-effects models for modelling resource selection. Journal of Applied
Ecology 46:590–599.
Kruskal, J.B. 1964 Nonmetric multidimensional scaling: a numerical method.
Psychometrika 29:115-129.
Kruuk, H. 1972. Surplus killing by carnivores. Journal of the Zoological Society of
London 166:233-244.
Kunkel K. and Pletscher D.H. 2001. Winter hunting patterns of wolves in and near
Glacier National Park, Montana. The Journal of Wildlife Management 65:520-530.
Lankin, V. 1997. Factors of diversity of domestic behaviour in sheep. Genetics Selection
Evolution 29:73–92.
Laporte, I. 2008. Behavioural effects of wolf presence: responses of cattle and elk.
Master Thesis. University of Calgary. 93 pp.
126
Laporte, I., Muhly, T.B., Pitt, J.A., Alexander, M. and Musiani, M. 2010. Effects of
wolves on elk and cattle behaviours: implications for livestock production and wolf
conservation. PLoS ONE 5(8): e11954. doi:10.1371/journal.pone.0011954
Laundré, J.W., Hernández, L. and Altendorf, K.B. 2001. Wolves, elk, and bison:
reestablishing the “landscape of fear” in Yellowstone National Park, U.S.A. Canadian
Journal of Zoology 79:1401–1409.
Lehmkuhler, J., Palmquist, G., Ruid, D., Willging, B. and Wydeven, A. 2007. Effects of
wolves and other predators on farms in Wisconsin: Beyond verified losses. Pub-ER658 2007, Wisconsin Department of Natural Resources, Wisconsin, U.S.A.
Lele, S.R. 2009. A new method for estimation of resource selection probability function.
Journal of Wildlife Management 73:122–127.
Licht, D.S., Millspaugh, J.J., Kunkel, K.E., Kochanny, C.O. and Peterson, R.O. 2010.
Using small populations of wolves for ecosystem restoration and stewardship.
BioScience 60:147-153.
Lima, S.L. 1998. Stress and decision making under the risk of predation: recent
developments from behavioral, reproductive, and ecological perspectives. Advances
in the Study of Behaviour 27:215-290.
Lima, S.L. and Dill L.M. 1990. Behavioral decisions made under the risk of predation: a
review and prospectus. Canadian Journal of Zoology 68:619–640.
Lind, J. and Cresswell, W. 2005. Determining the fitness consequences of antipredation
behaviour. Behavioral Ecology 16:945-956.
Lubchenco, J., Olson, A.M., Brubaker, L.B., Carpenter, S.R., Holland, M.M., Hubbell,
S.P., Levin, S.A., MacMahon, J.A., Matson, P.A., Melillo, J.M., Mooney, H.A.,
127
Peterson, C.H., Pulliam, H.R., Real, L.A., Regal, P.J. and Risser, P.G. 1991. The
sustainable biosphere initiative: an ecological research agenda: a report from the
ecological society of America. Ecology 72:371-412.
Lyon, L.J. 1979. Habitat effectiveness for elk as influenced by roads and cover. Journal
of Forestry 77:658-660.
Mackay, P., Zielinski, W.J., Long, R.A. and Ray, J.C. 2008. Noninvasive research and
carnivore conservation. In: R.A. Long, P. MacKay, W.J. Zielinski and J.C. Ray
(Editors), Noninvasive survey methods for carnivores Island Press, Washington,
D.C., pp. 1-7.
Manly, B.F.J., McDonald, L.L., Thomas, D.L., McDonald, T.L. and Erickson, W.P.
2002. Resource selection by animals: statistical analysis and design for field studies,
Second Edition. Nordrecht, The Netherlands: Kluwer, pp.240.
Manning, A.D., Gordon, I.J. and Ripple, W.J. 2009. Restoring landscapes of fear with
wolves in the Scottish Highlands. Biological Conservation 142:2314-2321.
Mather, P.M. 1976. Computational methods of multivariate analysis in physical
geography. J. Wiley and Sons, London, pp. 532.
McCune, B. and Grace, J.B. 2002. Analysis of ecological communities. MjM Software
Design, Gleneden Beach, Oregon, pp. 300.
McCune, B. and Mefford, M.J. 2006. PC-ORD, multivariate analysis of ecological data,
version 5. MjM Software, Gleneden Beach, Oregon, U.S.A.
McDermid, G.J., Hall, R.J., Sanchez-Azofeifa, G.A., Franklin, S.E., Stenhouse, G.B.,
Kobliuk, T. and LeDrew, E.F. 2009. Remote sensing and forest inventory for wildlife
habitat assessment. Forest Ecology and Management 257:2262–2269.
128
McGarigal, K. and Cushman, S.A. 2002. Comparative evaluation of experimental
approach to the study of habitat fragmentation effects. Ecological Applications
12:335-345.
McLoughlin, P.D., Morris, D.W., Fortin, D., Vander Wal, E. and Contasti, A.L. 2009.
Considering ecological dynamics in resource selection functions. Journal of Animal
Ecology 79:4-12.
Mech, L.D. 1970. The wolf: the ecology and behaviour of an endangered species.
University of Minnesota Press, Minneapolis, MN, 384 pp.
Mech, L.D. 1995. The challenge and opportunity of recovering wolf populations.
Conservation Biology 9:270-278.
Mech, L.D. and Barber, S.M. 2002. A critique of wildlife radio-tracking and its use in
national parks. U.S. National Park Service. Pp. 80.
Melis, C., Jedrzejewska, B., Apollonio, M., Barton, K.A., Jedrzejewski W, Linnell, J.D.,
Kojola, I., Kusak, J., Adamic, M., Ciuti, S., Delehan, I., Dykyy, I., Krapinec, K.,
Mattioli, L., Sagaydak, A., Samchuk, N., Schmidt, K., Shkvyrya, M., Sidorovich,
V.E., Zawadzka, B. and Zhyla, S. 2009. Predation has a greater impact in less
productive environments: variation in roe deer, Capreolus capreolus, population
density across Europe. Global Ecology and Biogeography 18:724–734.
Merrill, S.B. and Mech, L.D. 2003. The usefulness of GPS telemetry to study wolf
circadian and social activity. Wildlife Society Bulletin 31:947-960.
Mignon-Grasteau, S., Boissy, A., Bouix, J., Faure, J.M., Fisher, A.D., Hinch, G.N.,
Jensen, P., LeNeindre, P., Mormède, P., Prunet, P., Vandeputte, M. and Beaumont, C.
129
2005. Genetics of adaptation and domestication in livestock. Livestock Production
Science 93:3-14.
Miller, F.L., Gunn, A. and Broughton, E. 1985. Surplus killing as exemplified by wolf
predation on newborn caribou. Canadian Journal of Zoology 63:295-300.
Milton, S.J., Dean, W.R.J., du Plessis, M.A. and Siegfried, W.R. 1994. A conceptual
model of arid rangeland degradation. BioScience 44:70–76.
Minchin, P.R. 1987. An evaluation of the relative robustness of techniques for ecological
ordination. Vegetation 69:89–107.
Mitchell, J.E., Knight, R.L. and Camp, R.J. 2002. Landscape attributes of subdivided
ranches. Rangelands 24:3-9.
Mitchell, W.A. and. Lima, S.L. 2002. Predator-prey shell games: large-scale movement
and its implications for decision-making by prey. Oikos 99:249-259.
Montag, J. 2003. Compensation and predator conservation: Limitations of compensation.
Carnivore Damage Prevention News 6:2-6.
Morgantini, L.E. and Hudson, R.J. 1985. Changes in diets of wapiti during a hunting
season. Journal of Range Management 38:77-79.
Morgantini, L.E. and Hudson, R.J. 1989. Nutritional significance of wapiti migrations to
alpine ranges in western Alberta. Arctic and Alpine Research 21:288-295.
Muhly, T.B., Gates, C.C., Callaghan, C. and Musiani, M. 2010a. Predicting livestock
depredation by wolves in southwestern Alberta: implications for wolf management in
agricultural areas. In: Musiani M., L. Boitani, and P. Paquet (Editors). World of
Wolves: new perspectives on ecology, behaviour and management. University of
Calgary Press.
130
Muhly, T.B., Alexander, M., Boyce, M.S., Creasey, R., Hebblewhite, M., Paton, D., Pitt,
J.A. and Musiani, M. 2010b. Differential risk effects of wolves on wild versus domestic
prey have consequences for conservation. Oikos 119:1243-1254.
Musiani, M., Anwar, M. and Marceau, D. 2010. How humans shape wolf behaviour in
Banff and Kootenay National Parks, Canada. Ecological Modeling 221:2374-2387.
Musiani, M., Mamo, C., Boitani, L., Callaghan, C., Gates, C.C., Mattei, L., Visalberghi,
E., Breck, S. and Volpi, G. 2003. Wolf depredation trends and the use of fladry
barriers to protect livestock in western North America. Conservation Biology
17:1538-1547.
Musiani, M., Muhly, T., Gates, C. C., Callaghan, C., Smith, M.E. and Tosoni, E. 2005.
Seasonality and reoccurrence of depredation and wolf control in western North
America. Wildlife Society Bulletin 33:876-887.
Musiani, M. and Paquet P.C. 2004. The practices of wolf persecution, protection and
restoration in Canada and the USA. Bioscience 54:50-60.
National Park Service. 2008. Travel time cost surface model, Version 1.7. Fort Collins:
Rocky Mountain Network-National Park Service.
National Research Council of Canada. 2010. Sunrise/Sunset Calculator. Available:
http://www.nrc-cnrc.gc.ca/eng/services/hia/sunrise-sunset/angle-calculator.html.
Accessed 28 April 2010.
Naughton-Treves, L., Grossberg, R. and Treves, A. 2003. Paying for tolerance: the
impact of depredation and compensation payments on rural citizens’ attitudes toward
wolves. Conservation Biology 17:1500-1511.
131
Nelson, M.E. and Mech, L.D. 1986. Relationship between snow depth and gray wolf
predation on white-tailed deer. The Journal of Wildlife Management 50:471-474.
Niemeyer, C.C., Bangs, E.E., Fritts, S.H., Fontaine, J.A., Jimenez, M.D. and Brewster,
W.G. 1994. Wolf depredation management in relation to wolf recovery. In: W.S.
Halverson and A.C. Crabb (Editors), Proceedings of the Sixteenth Vertebrate Pest
Conference, University of California, Davis, California, pp. 57-60.
Northrup, J.M., Pitt, J., Muhly, T.B., Stenhouse, G.B., Musiani, M. and Boyce, M.S. (In
Review) Grizzly bears and traffic: linking human use to wildlife response to roads.
Journal of Wildlife Management.
Oakleaf, J.K., Mack, C. and Murray, D.L. 2003. Effects of wolves on livestock calf
survival and movements in Central Idaho. Journal of Wildlife Management 67:299306.
O'Brien, T.G., Kinnaird, M.F. and Wibisono, H.T. 2003. Crouching tigers, hidden prey:
Sumatran tiger and prey populations in a tropical forest landscape. Animal
Conservation 6:131-139.
Ormerod, S.J. 2002. Applied issues with predators and predation: editor’s introduction.
Journal of Applied Ecology 39:181–188.
Pace, M.L., Cole, J.J., Carpenter, S.R. and Kitchell, J.F. 1999. Trophic cascades revealed
in diverse ecosystems. Trends in Ecology and Evolution 14:483-488.
Paine, R.T. 1980. Food webs: linkage, interaction strength and community infrastructure.
Journal of Animal Ecology 49:666-685.
Palmer, M., Bernhardt, E., Chornesky, E., Collins, S. and Dobson, A. 2004. Ecology for a
crowded planet. Science 304:1251-1252.
132
Pechanec, J.F. and Pickford, G.D. 1937. A weight estimate method for the determination
of range or pasture production. Journal of the American Society of Agronomics
29:894-904.
Peek, J.M. 1980. Natural regulation of ungulates (what constitutes a real wilderness?).
Wildlife Society Bulletin 8:217-227.
Peterson, A.T. and Robins, C.R. 2003. Using ecological-niche modeling to predict barred
owl invasions with implications for spotted owl conservation. Conservation Biology
17:1161-1165.
Peterson, R.O. 1977. Wolf ecology and prey relationships on Isle Royale. U.S. National
Park Service Scientific Monograph Series 11, U.S. Government Printing Office,
Washington, D.C.
Pettorelli, N., Vik, J.O., Mysterud, A., Gaillard, J.-M., Tucker, C.J. and Stenseth, N.C.
2005. Using the satellite-derived NDVI to assess ecological responses to environmental
change. Trends in Ecology and Evolution 20:503-510.
Pigou, A.C. 1932. The Economics of Welfare, 4th Edition. Macmillan and Co., London,
U.K.
Polis, G.A. 1999. Why are parts of the world green? Multiple factors control productivity
and the distribution of biomass. Oikos 86:3-15.
Power, M.E. 1992. Top-down and bottom-up forces in food webs: do plants have
primacy? Ecology 73:733-746.
Preisler, H.K., Ager, A.A. and Wisdom, M.J. 2006. Statistical methods for analysing
responses of wildlife to human disturbance. Journal of Applied Ecology 43:164–172.
133
Preisser, E.L. Bolnick, D.I. and Benard, M.F. 2005. Scared to death? The effects of
intimidation and consumption in predator–prey interactions. Ecology 86:501–509.
Preisser, E.L. and Bolnick, D.I. 2008. The many faces of fear: comparing the pathways
and impacts of nonconsumptive predator effects on prey populations. PLoS ONE
3:e2465.
Price, E.O. 1984. Behavioural aspects of animal domestication. Quarterly Review of
Biology 59:1-32.
Rabe-Hesketh, S. Skrondal, A. and Pickles, A. 2005. Maximum likelihood estimation of
limited and discrete dependent variable models with nested random effects. Journal of
Econometrics 128:301–323.
Rasker, R. and Hackman, A. 1996. Economic development and the conservation of large
carnivores. Conservation Biology 10:991-1002.
Ray, J.C., Redford, K.H., Steneck, R.S. and Berger, J. 2005. Large carnivores and the
conservation of biodiversity. Washington, DC, Island Press, 526 pp.
Riley, S.J., DeGloria, S.D. and Elliot, R. 1999. A terrain ruggedness index that quantifies
topographic heterogeneity. Intermountain Journal of Sciences 5:1–4.
Ripple, W.J. and Beschta, R.L. 2006. Linking a cougar decline, trophic cascade, and
catastrophic regime shift in Zion National Park. Biological Conservation 133:397–408.
Ripple, W.J., Larsen, E.J., Renkin, R.A. and Smith, D.W. 2001. Trophic cascades among
wolves, elk and aspen on Yellowstone National Park’s northern range. Biological
Conservation 102:227–234.
Roberts, G. 1996. Why individual vigilance declines as group size increases. Animal
Behaviour 51:1077-1086.
134
Romeyer, A. and Bouissou, M.F. 1992. Assessment of fear reactions in domestic sheep
and influence of breed and rearing conditions. Applied Animal Behaviour Science
34:93–119.
Roth II, T.C. and Lima, S.L. 2007. Use of prey hotspots by an avian predator: purposeful
unpredictability? The American Naturalist 169:264–273.
Rovero, F. and Marshall, A.R. 2009. Camera trapping photographic rate as an index of
density in forest ungulates. Journal of Applied Ecology 46:1011–1017.
Rowcliffe, J.M., Field, J., Turvey, S.T. and Carbone, C. 2008. Estimating animal density
using camera traps without the need for individual recognition. Journal of Applied
Ecology 45:1228-1236.
Sarchet, B.A. 2005. Survey and analysis of Wyoming livestock producers. Masters thesis,
University of Wyoming, Laramie, WY.
Sawyer, H., Nielson, R.M., Lindzey, F. and Mcdonald, L.L. 2006. Winter habitat
selection of mule deer before and during development of a natural gas field. Journal
of Wildlife Management 70:396–403.
Schmitz, O.J. 1994. Resource edibility and trophic exploitation in an old-field food web.
Proceedings of the National Academy of Science 91:5364-5367.
Schmitz, O.J. 1998. Direct and indirect effects of predation and predation risk in old-field
interaction webs. The American Naturalist 151:327-342.
Schmitz, O.J., Beckerman, A.P. and O'Brien, K.M. 1997. Behaviorally mediated trophic
cascades: Effects of predation risk on food web interactions. Ecology 78:1388-1399.
Schmitz, O.J., Krivan, V. and Ovadia, O. 2004. Trophic cascades: the primacy of traitmediated indirect interactions. Ecology Letters 7:153–163.
135
Schmitz, O.J. and Suttle, K.B. 2001. Effects of top predator species on direct and indirect
interactions in a food web. Ecology 82:2072-2081.
Schmitz, O.J., Beckerman, A.P. and O'Brien, K.M. 1997. Behaviorally mediated trophic
cascades: effects of predation risk on food web interactions. Ecology 78:1388-1399.
Schultz, R.D. and Bailey, J.A. 1978. Responses of national park elk to human activity.
Journal of Wildlife Management 42:91-100.
Seaman, D.E. and Powell, R.A. 1996. An evaluation of the accuracy of kernel density
estimators for home range analysis. Ecology 77:2075-2085.
Sengupta, S. and Osgood, D.E. 2003. The value of remoteness: a hedonic estimation of
ranchette prices. Ecological Economics 44:91-103.
Shivik, J.A. 2006. Tools for the edge: What's new for conserving carnivores. BioScience
56:253–259.
Short, J., Kinnear, J. E. and Robley, A. 2002. Surplus killing by introduced predators in
Australia evidence for ineffective anti-predator adaptations in native prey species?
Biological Conservation 103:283-301.
Sih, A. 1984. The behavioral response race between predator and prey. The American
Naturalist 123:143-150.
Sih, A. 2005. Predator-prey space use as an emergent outcome of a behavioral response
race. In: P. Barbosa and I. Castellanos (Editors), Ecology of predator-prey interactions,
Oxford University Press, Oxford, UK., pp. 240-255.
Sjöberg, S. 1980. Zooplankton feeding and queuing theory. Ecological Modelling,
10:215-225.
136
Skovlin, J.M., Zager, P. and Johnson, P.K. 2002. Elk habitat selection and evaluation. In:
D.E. Toweill and J.W. Thomas (Editors), North American elk: Ecology and
management. Smithsonian Institution Press, pp. 531–555.
Skrondal, A. and Rabe-Hesketh, S. 2004. Generalized latent variable modeling:
multilevel, longitudinal, and structural equation models. Chapman and Hall.
Smith BD (2007) The ultimate ecosystem engineers. Science 315:1797-1798.
Sokal, R.S. and Rohlf, J.R. 2000. Biometry. 3rd edition. W. H. Freeman and Company,
New York, NY, pp. 859.
Stankowich, T. 2008 Ungulate flight responses to human disturbance: A review and
meta-analysis. Biological Conservation 141:2159–2173.
StataCorp. 2009. Stata/SE 10.1 for Windows. Stata Corporation. College Station, Texas.
Stewart, K.M., Bowyer, R.T., Kie, J.G., Cimon, N.J. and Johnson, B.K. 2002.
Temporospatial distributions of elk, mule deer, and cattle: resource partitioning and
competitive displacement. Journal of Mammalogy 83:229–244.
Stuart, C.T. 1986. The incidence of surplus killing by Panthera pardus and Felis caracal
in Cape province, South Africa. Mammalia 50:556-558.
Talbert, C.B., Knight, R.L. and Mitchell, J.E. 2007. Private ranchlands and public land
grazing in the southern Rocky Mountains. Rangelands 29:5–8.
Telleria, J.L. and Saez-Royuela, C. 1989. Ecología de una población ibérica de lobos
(Canis lupus), Doñana. Acta Vertebrata 16:105-122.
Theobald, D.M., Miller, J.R. and Hobbs, N.T. 1997. Estimating the cumulative effects of
development on wildlife habitat. Landscape and Urban Planning 39:25-36.
137
Theuerkauf, J., Rouys, S. and Jedrzejewski, W. 2003. Selection of den, rendezvous, and
resting sites by wolves in the Bialowieza Forest, Poland. Canadian Journal of
Zoology 81:163–167.
Tobler, M.W., Carrillo-Percastegui, S.E., Pitman, R.L., Mares, R. and Powell, G. 2008.
An evaluation of camera traps for inventorying large- and medium-sized terrestrial
rainforest mammals. Animal Conservation 11:169–178.
Treves, A. and Karanth, K.U. 2003. Human-carnivore conflict: local solutions with
global applications. Conservation Biology 17:1489–1490.
Treves, A., Mwima, P., Plumptre, A.J. and Isoke, S. 2010. Camera-trapping forest–
woodland wildlife of western Uganda reveals how gregariousness biases estimates of
relative abundance and distribution. Biological Conservation 143:521–528.
Trombulak, S.C. and Frissell, C.A. 2000. Review of ecological effects of roads on
terrestrial and aquatic communities. Conservation Biology 14:18-30.
Tucker, C.J. 1980. A critical review of remote sensing and other methods for nondestructive estimation of standing crop biomass. Grass and Forage Science 35:177182.
Turner II, B.L., Lambin, E.F. and Reenberg, A. 2007. The emergence of land change
science for global environmental change and sustainability. Proceedings of the
National Academy of Science 104:20666–20671.
Tylianakis, J.M., Didham, R.K., Bascompte, J. and Wardle, D.A. 2008. Global change
and species interactions in terrestrial ecosystems. Ecology Letters 11:1351–1363.
138
United States Department of Agriculture (USDA). 1993. Meat animals production,
agricultural disposition, and income, final estimates, 1988-1992. National
Agricultural Statistics Service. Statistical Bulletin Number 907, Washington, D.C.
United States Department of Agriculture (USDA). 1998. Meat animals production,
agricultural disposition, and income, final estimates, 1993-1997. National
Agricultural Statistics Service. Statistical Bulletin Number 959a, Washington, D.C.
United States Department of Agriculture (USDA). 1999. Agricultural land values and
cash rents, final estimates,1994-98. Statistical Bulletin Number 957. National
Agricultural Statistics Service, Washington, D.C.
United States Department of Agriculture (USDA). 2003. Livestock slaughter, 2002
summary. National Agricultural Statistics Service, Statistical Bulletin Number
MtAn1-2-1(03)a, Washington, D.C.
United States Department of Agriculture (USDA). 2004a. Meat animals production,
agricultural disposition, and income, final estimates, 1998-2002. National
Agricultural Statistics Service, Statistical Bulletin Number 995(04), Washington,
D.C.
United States Department of Agriculture (USDA). 2004b. Agricultural land values and
cash rents, final estimates, 1999-2003. Statistical Bulletin Number 993. National
Agricultural Statistics Service, Washington, D.C.
United States Department of Agriculture (USDA). 2008. Agricultural prices. National
Agricultural Statistics Service. Washington, D.C.
United States Fish and Wildlife Service (USFWS). 2008. Endangered and threatened
wildlife and plants; final rule designating the northern Rocky Mountain population of
139
gray wolf as a distinct population segment and removing this distinct population
segment from the federal list of endangered and threatened wildlife; final rule.
Federal Register/, ol. 73, No. 39, U.S. Department of the Interior, Fish and Wildlife
Service, Washington, D.C.
Vitousek, P.M., Mooney, H.A., Lubchenco, J. and Melillo, J.M. 1997. Human
domination of Earth's ecosystems. Science 277:494-499.
Wagner, K.K., Schmidt, R.H. and Conover, M.R. 1997. Compensation programs for
wildlife damage in North America. Wildlife Society Bulletin 25:312-319.
Werner, E.E. and Peacor, S.D. 2003. A review of trait-mediated indirect interactions in
ecological communities. Ecology 84:1083–1100.
Whittington, J., Cassady St. Clair, C. and Mercer G. 2005. Responses of wolves to roads
and trails in mountain valleys. Ecological Applications 15:543-553.
Willems, E.P. and Hill, R.A. 2009. Predator-specific landscapes of fear and resource
distribution: effects on spatial range use. Ecology 90:546-555.
Wolf, M., Frair, J., Merrill, E. and Turchin, P. 2009. The attraction of the known: the
importance of spatial familiarity in habitat selection in wapiti Cervus elaphus.
Ecography 32:401-410.
Wood-Gush, D. and Duncan, I. 1976. Some behavioural observations on domestic fowl
in the wild. Applied Animal Ethology 2:255-260.
Woodroffe, R. 2000. Predators and people: using human densities to interpret declines of
large carnivores. Animal Conservation 3:165–173.
Woodroffe, R. and Ginsberg, J.R. 1998. Edge effects and the extinction of populations
inside protected areas. Science 280:2126–2128.
140
Wuerthner, G. 1994. Subdivisions versus agriculture. Conservation Biology 8:905-908.
Young, F.W. and Hamer, R.M. 1987. Multidimensional Scaling: history, theory, and
applications. Lawrence Erlbaum Associates, Hillsdale, NJ, pp. 318.
Zabel, A. and Holm-Müller, K. 2008. Conservation performance payments for carnivore
conservation in Sweden. Conservation Biology 22:247–251.
141
APPENDIX A. RESOURCE SELECTION FUNCTION COEFFICIENTS AND STANDARD ERRORS
FOR ELK AND CATTLE BEFORE, DURING AND AFTER WOLF VISITS TO HOME RANGES AND
PASTURES, RESPECTIVELY, CALCULATED FROM GENERALIZED LINEAR MIXED MODELS.
Table A1. Elk resource selection function coefficients (β) and standard errors in periods
before, during and after wolf visits to elk home ranges calculated from two Generalized
Linear Mixed Models (GLMMs) with distance to forest cover or food habitat quality as
conditional coefficient, respectively.
Random Coefficient =
Distance to Forest Cover
Covariate
Fixed Effects
Terrain ruggedness
Distance to road or trail
(km)
Distance to forest cover
(km)
Food Habitat Quality
Terrain ruggedness *
Food Habitat Quality
Distance to road or trail
(km) * Food Habitat
Quality
Distance to forest cover
(km) * Food Habitat
Quality
Random Coefficient =
Food Habitat Quality
Before
During
After
Before
During
After
-0.026*
-0.030*
-0.022*
-0.032**
-0.041***
-0.023**
SE
0.012
0.012
0.009
0.013
0.012
0.009
β
-0.008
0.094
0.063
0.087
0.576
0.150
SE
0.263
0.250
0.186
0.309
0.296
0.183
5.469***
0.329
0.803
2.576***
1.768***
1.353***
SE
1.034
0.533
0.434
0.867
0.315
0.222
β
β
β
0.559*
0.156
0.426*
1.025**
0.495
0.574*
SE
0.281
0.248
0.195
0.398
0.402
0.240
β
0.027*
0.045***
0.021*
0.031*
0.059***
0.024**
SE
0.013
0.012
0.009
0.014
0.013
0.009
β
-0.450
-0.494*
-0.457*
-0.393
-0.968***
-0.553***
SE
0.279
0.246
0.166
0.329
0.297
0.170
-5.286***
-0.205
-0.253**
-2.365**
-0.613**
-0.357***
0.985
0.115
0.089
0.873
0.231
0.100
β
SE
142
Random Effects
γk(visit)
γ1k(visit) x1jk
Intercept
0.930
1.259
0.745
1.943
2.253
0.986
SE
0.209
0.275
0.174
0.819
0.835
0.347
Coefficient
8.648
11.297
7.341
1.341
2.251
0.303
SE
2.122
2.952
2.071
0.752
0.886
0.215
-2.652
-2.631
-1.625
-1.592
-1.890
-0.546
0.614
0.759
0.499
0.781
0.818
0.284
-0.935
-0.698
-0.695
-0.987
-0.840
-1.000
Covariance
SE
Correlation
* = significant P < 0.05
** = significant P < 0.01
*** = significant P < 0.001
143
Table A2. Cattle resource selection function coefficients (β) and standard errors in periods
before, during and after wolf visits to cattle pastures calculated from two Generalized
Linear Mixed Models (GLMMs) with distance to forest cover or food habitat quality as
conditional coefficient, respectively.
Random Coefficient =
Distance to Forest Cover
Covariate
Fixed Effects
Terrain ruggedness
β
SE
Distance to road or trail
(km)
β
SE
Distance to forest cover
(km)
β
SE
Food Habitat Quality
β
SE
Terrain ruggedness *
Food Habitat Quality
β
SE
Distance to road or trail
(km) * Food Habitat
Quality
β
SE
Distance to forest cover
(km) * Food Habitat
Quality
β
SE
Random Coefficient =
Food Habitat Quality
Before
During
After
Before
During
After
-0.040***
-0.047***
-0.021***
-0.037***
-0.038***
-0.023***
0.003
0.004
0.003
0.003
0.004
0.003
1.059***
1.150***
-1.725***
1.032***
1.183***
-1.731***
0.123
0.142
0.136
0.123
0.139
0.138
48.600***
52.516***
50.774***
50.765***
55.386***
51.821***
6.009
7.911
6.898
7.108
8.369
7.138
0.544***
0.585***
-0.189**
0.696***
1.130***
-0.381***
0.081
0.095
0.068
0.083
0.010
0.076
0.043***
0.039***
0.023***
0.038***
0.027***
0.024***
0.004
0.005
0.003
0.004
0.005
0.003
-1.709***
-0.276
2.359***
-1.626***
-0.312***
2.328***
0.145
0.160
0.151
0.145
0.158
0.153
44.747***
50.833***
50.993***
-46.941***
-53.221***
-49.741***
6.008
7.912
6.901
7.109
8.370
7.139
144
Random Effects
γjk(cow)
Intercept
γk(visit)
γ1jk
(visit)
x1ijk
0.004
0.018
0.241
0.000
0.120
0.232
SE
0.003
0.008
0.014
0.000
0.021
0.020
Intercept
0.069
0.195
0.071
0.491
1.398
0.410
SE
0.008
0.016
0.005
0.053
0.146
0.021
Coefficient
6.916
8.178
20.902
0.581
0.437
0.336
SE
0.627
0.675
1.289
0.059
0.036
0.022
-0.486
-0.151
-0.341
-0.518
-0.283
-0.352
0.046
0.063
0.041
0.055
0.090
0.021
-0.703
-0.119
-0.280
-0.970
-0.362
-0.947
Covariance
SE
Correlation
* = significant P < 0.05
** = significant P < 0.01
*** = significant P < 0.001
145
APPENDIX B: HUMAN DISTRIBTUION MODEL FOR SOUTHWEST ALBERTA, CANADA
A human spatial distribution model (Apps et al. 2004) was calculated to provide a
relative index of human density. It was based on the travel time required to access any
point along existing road and trail networks from human population centers, given typical
travel speeds on each road and trail type and a decay exponent of –1.45 based on typical
behaviour of travelers. To calculate travel times I used the Travel Time Cost Surface
Extension (National Park Service 2008) in ArcGIS 9.2 (ESRI 2009). Travel speeds were
obtained from posted road speed limits and typical All-Terrain Vehicle (ATV) speeds on
trails (Durbin 2004). Population centers included all residential areas with a population
>100 people within the study area. I obtained GIS data on roads and trails from the
Government of Alberta and corrected errors in the data by overlaying it on a digital air
photo with 5 m spatial resolution and manually editing observed errors.
A significant linear relationship was found between the human access index and
actual number of humans counted on roads and trails in the study during the day
(R2=0.61, F=197.78, p<0.001) and night (R2=0.57, F=169.85, p<0.001) indicating I had
predictive models of actual human daytime and nighttime counts. Data were logtransformed to make them normal (tested using skewness and kurtosis tests).
146
(a)
(b)
Figure B1. Linear relationship between the human density index and actual counts of
humans on roads and trails during the day (a), and night (b) in southwest Alberta, Canada
in 2008. Each linear relationship was significant: (a) R2=0.605, F=197.78, p<0.001; (b)
R2=0.573, F=169.85, p<0.001. Count data are from 43 trail cameras and 43 road
counters. Data were log transformed to make them normal.
147
APPENDIX C: ELK, CATTLE AND WOLF RESOURCE SELECTION FUNCTION (RSF) MODELS
FOR SOUTHWEST ALBERTA, CANADA.
Resource selection functions (RSFs) were estimated using logistic regression,
where resource units at telemetry locations are compared with resource units at
“available” locations (Johnson t al. 2006; Lele 2009), following RSF ‘sampling protocol
A’ (Manly et al. 2002). Available resource units were sampled at random locations in
each animal home range (for wolves and elk) or pasture (for cattle). I estimated wolf and
elk home ranges using a 95% kernel density estimator (Seaman and Powell 1996) of
telemetry location data, with a smoothing parameter (h=3 km and 1 km, respectively) that
I selected from a comparison of several different smoothing factors because it effectively
portrayed wolf and elk distribution in the study area. Habitats available to cattle were
defined by fenced pasture boundaries.
Resource covariates that I included in my model were: density of roads and trails,
slope, vegetation cover and distance to water. I also included squared covariates for
road/trail density and slope to account for non-linear relationships. Vegetation data
consisted of 13 vegetation cover types and was obtained from a 30-m2 spatial resolution
GIS map derived from Landsat data (McDermid et al. 2009). Overall, classification
accuracy of the vegetation cover map was 80%. Slope was calculated from a 30-m2
spatial resolution digital elevation model (DEM). I calculated road and trail density from
a GIS dataset of roads and trails obtained from the government of Alberta. All GIS work
was completed using ArcGIS 9.2 (ESRI 2009), including Spatial Analyst and Hawth’s
Analysis Tools (Beyer 2004) extensions.
2
0.0478
0.05
Barren Ground
Intercept
-188.9203
0.02
0.02
<0.01
0.02
0.02
0.02
0.02
Wolf
72.7008
Dropped
-0.01
Agricultural Field
Snow/Ice
0.05
0.05
Regenerating Forest
Herbaceous
0.07
Broadleaf Forest
Shrub
0.05
0.06
Mixed Forest
0.02
0.02
0.05
0.06
Open Conifer Forest
0.02
0.05
Dense Conifer Forest
Moderately Closed Canopy Conifer
Forest
Dropped
0.0220
0.0022
0.0716
Open Wetland
0.0664
Distance to Water (m)
SE
0.1946
Dropped
-0.0104
Slope2
Day
Treed Wetland
0.2863
-0.1187
0.1176
β
Slope (°)
Road and Trail Density
2
Road and Trail Density (km/km )
Covariate
et al. 2010).
0.0551
-125.9864
56.0790
0.01
<0.01
0.02
0.02
0.01
0.02
0.02
0.02
0.02
0.02
<0.01
0.0168
0.0022
0.0468
Dropped
0.03
-0.01
0.04
0.04
0.05
0.04
0.04
0.04
0.04
0.04
SE
0.1907
Dropped
<-0.01
0.0381
-0.0069
0.0634
-0.1646
0.3841
β
Night
0.0144
<0.0001
0.0003
0.0098
0.57
-2.5687
-0.68
0.26
0.55
1.33
1.03
0.12
1.32
0.74
1.52
0.5722
0.35
0.35
0.08
0.51
0.53
0.06
0.52
0.51
0.52
0.52
0.52
Dropped
0.12
SE
0.0700
Dropped
<-0.0001
-0.0041
0.1257
-0.0310
-0.0948
β
Day
β
0.0152
<0.0001
0.0002
0.0098
0.6555
0.46
0.48
0.11
0.59
0.59
0.07
0.57
0.58
0.59
0.58
0.54
Dropped
-4.1847
1.62
1.24
0.67
2.04
1.70
-0.38
2.11
1.44
2.290
1.15
0.81
SE
0.0773
Night
Dropped
<-0.0001
-0.0039
0.1219
-0.0355
-0.1074
Elk
0.0232
-0.2359
0.23
0.59
0.24
0.22
0.28
0.22
0.23
0.7997
Dropped
-2.31
Dropped
0.90
0.85
Dropped
1.09
0.46
0.57
-0.07
0.27
<0.0001
0.0009
0.0188
Dropped
-0.84
SE
0.1186
Dropped
<-0.0001
-0.0016
-0.0702
-0.0230
0.2905
β
Day
β
0.0273
<0.0001
0.0016
0.0327
0.7578
Dropped
0.47
Dropped
0.24
0.24
Dropped
0.26
0.25
0.28
0.26
0.30
Dropped
-1.9132
-1.75
1.47
1.27
0.62
0.51
1.29
-0.30
-0.79
SE
0.1236
Night
Dropped
<-0.0001
-0.0085
0.0913
-0.0299
0.2965
Cattle
the day (sunrise to sunset) and night (sunset to sunrise) in southwest Alberta, Canada, calculated using the two-stage method (Fieberg
Table C1. Coefficients of population-level resource selection function (RSF) models of radio-collared wolves, elk and cattle during
148
149
APPENDIX D: ELK, CATTLE AND WOLF RESOURCE SELECTION FUNCTION MODEL
VALIDATION
I validated my population-averaged wolf and elk RSF models using k-fold cross
validation (Boyce et al. 2002). Spearman correlations were calculated between RSF ranks
and area-adjusted frequencies from a withheld sub-sample of data. The 5-fold cross
validation used location data from all individual animals, withholding 20% of the data for
each iteration. I further tested the validity of my model by calculating a linear regression
between observed frequency and expected RSF scores and assessing the fit (Johnson et
al. 2006). Similarly I validated the cattle RSF using a 2-pasture cross validation, where
separate cattle RSFs were produced for each pasture and a Spearman rank correlation and
linear regression between the two models was calculated in each pasture. The validated
population-level RSFs served as predictive models of each species’ daytime and
nighttime distribution on the landscape.
Mean
R
2
Mean
Spearman rs
Group
1
2
3
4
5
1
2
3
4
5
0.941
0.863
0.769
0.874
0.857
0.839
1.000
0.988
0.964
0.952
0.994
0.979
Day
Wolf
0.914
0.794
0.761
0.841
0.839
0.884
1.000
0.976
0.988
0.988
0.951
0.981
Night
0.948
0.945
0.945
0.949
0.946
0.943
1.000
1.000
1.000
1.000
1.000
1.000
Day
Elk
0.925
0.911
0.938
0.926
0.924
0.920
1.000
1.000
1.000
1.000
1.000
1.000
Night
Bob Creek - Porcupine Hills
Porcupine Hills - Bob Creek
Bob Creek - Porcupine Hills
Porcupine Hills - Bob Creek
Pasture
daytime (sunrise to sunset) and nighttime (sunset to sunrise) resource selection functions (RSFs).
0.340
0.559
0.122
0.672
0.726
0.618
Day
Cattle
0.272
0.519
0.025
0.704
0.808
0.599
Night
Table D1. Spearman correlations and linear regression R2 values from 5-fold and pasture cross validation of wolf, elk and cattle
150
151
APPENDIX E: FORAGE UTILIZATION
Actual forage utilization was calculated from the difference in aboveground
phytomass of grasses and forbs clipped within paired caged and uncaged sub-plots (both
were 1.2 m x 1.2 m in size). Samples were clipped to an l-cm stubble height from a
0.5 m2 area within the plots, dried at 70°C for 48 hours and weighed to the nearest gram
(Irving et al. 1995). Percent utilization was calculated from the difference between caged
and un-caged dried biomass divided by the caged biomass and multiplied by 100 (Irving
et al. 1995). After the 2007 field season, I compared the clipped biomass utilization
values to the visual estimates of utilization in a linear regression to asses our ability to
predict visually utilization. I identified a significant linear relationship (R2=0.255,
F=47.50, p<0.001) and concluded that visual estimates were useful measures of
vegetation utilization.
152
Figure E1. Linear relationship between actual percent herbaceous vegetation biomass
utilization and a visual estimate of the percent vegetation biomass utilization by
herbivores at 40 vegetation plots in southwest Alberta, Canada. The linear relationship
was significant (R2= 0.255, F=47.50, p<0.001), which indicates that visual estimates of
vegetation utilization were accurate measures of actual utilization.
153
APPENDIX F: STRUCTURAL EQUATION MODELS INDICATING SPECIES RELATIONSHIPS
WITHIN HIGH-HUMAN AND LOW-HUMAN USE WOLF HOME RANGES.
I produced structural equation models using species distribution data only within
the home range boundaries of two wolf packs characterized as having the highest and
lowest average human counts on roads and trails within the home range. The purpose was
to evaluate whether the relative strength and type (positive or negative) of pathways
between species was different within wolf home ranges depending on the density of
humans. I sampled each species distribution model at 30-m pixels along roads and trails
within each wolf home range (n1 =65,559; n2 =73,559). I also sampled species
distribution at vegetation utilization plots located within each of the wolf home ranges
(n1=17; n2=80) to test for effects of wolves, elk and cattle on forage utilization.
Elk
-0.138
0.384
0.078
Forage
Quality and
Quantity
Wolf
0.365
GFI=1.000, AGFI=0.995, RMR=0.005, AIC=609.982
Cattle
0.452
0.436
-0.102
High-Human Density (Average 49 humans/day) Wolf Home Range During the Day
Human
154
Elk
-0.146
0.327
0.126
Forage
Quality and
Quantity
0.285
Wolf
0.385
0.086
GFI=0.998, AGFI=0.983, RMR=0.015, AIC=401.113
Cattle
0.452
0.378
High-Human Density (Average 6 humans/day) Wolf Home Range During the Night
Human
155
Elk
0.021
0.350
0.030
Forage
Quality and
Quantity
0.321
Wolf
0.494
0.140
GFI=1.000, AGFI=1.000, RMR=0.001, AIC=29.340
Cattle
0.358
0.162
-0.097
Low-Human Density (Average 10 humans/day) Wolf Home Range During the Day
Human
156
Elk
0.031
0.282
0.136
Forage
Quality and
Quantity
0.217
Wolf
0.463
0.066
GFI=0.994, AGFI=0.915, RMR=0.023, AIC=966.130
Cattle
0.262
0.245
0.173
Low-Human Density (Average 1 human/night) Wolf Home Range During the Night
Human
157
represented by solid lines. Human influences are represented by dashed-dotted lines.
Arrows indicate the pathway direction and thickness is proportional to coefficient value. Consumer-resource interactions are
human counts on roads and trails in southwest Alberta, Canada. Pathways were constructed using structural equation modeling (SEM).
quantity during the daytime and nighttime within two actual wolf home ranges with the highest (top) and lowest (bottom) average
Figure F1. Food web model illustrating the pathways between the distribution of humans, wolves, elk, cattle, and forage quality and
158