H7 .L n. 4": y. - I‘A,.P'.

‘ . .4

4b.. 4
I :1:
. vrmhrfh

52‘13.
a... ‘|!\ I

3.3:,
, . . lug

.(31 ‘4

a
.- m.
.r .96“

9“...
... .
71‘; it
s
,-

ufi! 3.3.!

, hfddwvfldfl3i. V
51:... u}; 11 .r 541$. . 1
.10 $51 .I
.23.", unilau . . x
I, o
1.. 3%“.
V . if.
is

3..., .
2... .
15““.3....$.
f

.l V .. ‘ '1 l\1....\.
ski-WK h ‘ u l.s.€.. $.42 :311
J {#be . 2.! llxgxux

.
2C

 

 

100?

This is to certify that the
dissertation entitled

ANTHROPOGENIC INFLUENCES ON THE BEHAVIOR OF
LARGE CARNIVORES IN THE NORTHERN SERENGETI
ECOSYSTEM

presented by

JOSEPH MARK KOLOWSKI

has been accepted towards fulfillment
of the requirements for the

PhD. degree in Zoology

 

 

Major Professor’s Sfiature
Y/ag/fi-

Date

 

MSU is an affirmative-action, equal-opportunity employer

. -.._.._ _._.-.-..._._._.—-—._A—.—-—.—-.—.—-—._o-t-—-—a---.—

.._.—-—o—o—.—I—o-o-n---o-n—-—c—-—-—.¢ '-

 

 

LIBRARY

Michigan State

University

 

 

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

 

DATEDUE

DAIEDUE

DATEDUE

 

1mm W

 

 

 

 

 

 

 

 

 

 

 

 

 

6/07 p:/CIRC/Date0ue.indd-p.1

 

 

 

ANTHROPOGENIC INFLUENCES ON THE BEHAVIOR OF LARGE
CARNIVORES IN THE NORTHERN SERENGETI ECOSYSTEM

By

Joseph Mark Kolowski

A DISSERTATION

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

DOCTOR OF PHILOSOPHY

Department of Zoology and Program in Ecology, Evolutionary Biology and
Behavior

2007

ABSTRACT

ANTHROPOGENIC INFLUENCES ON THE BEHAVIOR OF LARGE
CARNIVORES IN THE NORTHERN SERENGETI ECOSYSTEM

By
Joseph Mark Kolowski

Most large mammalian carnivores are in global decline, largely due to
habitat loss and their involvement in livestock depredation. Increasingly, large
carnivores are forced to adjust to living in landscapes characterized by human
activity and disturbance. It is therefore critical to understand both the extent to
which carnivores can adjust to human activities and the factors influencing
human tolerance of their presence. Here I document the activity patterns and
space use of three social groups, or clans, of spotted hyenas (Crocuta crocuta) in
the Masai Mara National Reserve, Kenya that vary in exposure to humans, and
investigate the degree to which these behaviors are influenced by human activity,
primarily in the form of livestock grazing. I also investigate the issue of local
tolerance of carnivores with an intensive study of livestock depredation.

Hyenas were active during 31.5 :I: 2.7% of the 24-h period, and 96.2 4.-
0.9% of all activity occurred from 1800—0900 h. Male spotted hyenas tended to
be more active, and exhibit higher movement rates than females. Female
hyenas in territories with daily livestock grazing showed lower activity and den
use than hyenas in an undisturbed territory during the times of day when
livestock grazing coincided with potential hyena activity.

Space use patterns of hyenas with no exposure to livestock grazing were

influenced by the location of the communal den, and the distribution of prey,

vegetation types, and water features within their territory. Relative to this clan,
hyenas exposed to livestock grazing showed a stronger avoidance of open grass
plains and a weaker association with prey resources and den location. However,
the distribution of livestock did not directly influence hyena space use patterns,
indicating that increased use of vegetative cover by hyenas may be an important
behavioral shift allowing temporary coexistence with livestock and their
herdsmen.

Hyena home range size, core area size, and core area location were
influenced by the presence of a refuse pit at the edge of one study clan’s
territory. The most common rank group utilizing the pit was low-ranking females,
and regular pit users were more likely to be found near the pit during times of
relative prey scarcity. These results indicate that human refuse at pastoral
villages may increase hyena use of these environments, and that this use may
vary on a seasonal and individual level.

Spotted hyenas, leopards (Panthera pardus) and lions (Panthera Ieo)
were responsible for 53%, 32%, and 15% of attacks on livestock, respectively,
that were documented along a Reserve border. Monthly depredation frequency
was correlated positively with rainfall and negatively with natural prey
abundance. The spatial location and size of local villages, and the fence type
used at livestock enclosures influenced the vulnerability of these locations to
livestock losses, with leopards and hyenas showing clear differences in selection
of attack locations. Overall, improved fences, more watch dogs, and high levels

of human activity were not associated with lower livestock losses to predators.

ACKNOWLEDGEMENTS

The research presented here was described in Animal Research Protocol
No. 05/05-064-00, approved most recently on 3 May 2006 by the All University
Committee on Animal Use and Care at Michigan State University (MSU).
Throughout my tenure the Mara Hyena Project has been supported by NSF
grants IBN0113170, IBN0343381, and IOBOG18022 to Kay Holekamp.
Throughout this program I personally received logistical and financial support
from the College of Natural Sciences, the Department of Zoology, and the
Ecology, Evolutionary Biology and Behavior program at MSU. In particular, the
friendly and incredibly helpful staff of the Zoology Department office made
navigating the waters of the MSU system manageable, and their assistance was
always greatly appreciated.

I thank the Office of the President of Kenya for allowing us to conduct this
research. I also thank the Kenya Wildlife Service, the Narok County Council and
the Senior Warden of the Masai Mara National Reserve for their cooperation and
assistance. I also thank the staff and management of Mara lntrepids Lodge for
their support, logistical and otherwise, without which this study would not have
been possible.

This dissertation was greatly improved by the insight and advice of my
graduate committee, which included Joe Messina, Scott Winterstein, Laura
Smale and my advisor Kay Holekamp. The logistical and financial support of Dr.
Holekamp made it possible for me to live and work in Kenya for almost 3 years,

allowing me to fulfill a nearly lifelong dream. I simply can’t thank her enough for

making that possible. In addition, her support, guidance, and attention throughout
this entire dissertation project was constant, it was critical to my success in
completing this project, and it did not go unappreciated.

ln spending close to 3 years in Kenya working on this project, I formed
many friendships that made living away from home for so long not only bearable,
but immensely enjoyable. Special thanks goes to the staff at Fisi Camp: John
Keshe, James Kerembe, Moses Sairowa, and our askaris Stephen and Lesingo
for keeping us safe, well-fed, and in good-spirits. In particular I thank John Keshe
for his warmth and friendship throughout my time in Kenya. I would surely still be
stuck on the Mara plains without the rescue efforts of Andrew Peart and his
balloon crew, and the staff and guides of Base Camp who never hesitated to
offer their help when needed. The work presented in Chapter 4 would not have
been possible without the efforts of my livestock-conflict scouts: P. Mkurrana, D.
Naurori, and D. Kasoe, the additional assistance of P. Naurori. Additionally, that
research relied almost completely on the cooperation and participation of
residents of our study area on the Koyake and Siana group ranches who
provided their support, assistance and insight.

I would like to thank all the current and previous members of the
Holekamp Lab, both graduate students, undergraduates, and field assistants that
have been a part of the enormous effort required to keep the Mara Hyena Project
going. Although many of these people, too numerous to name, have directly
assisted me in one way or another on my way toward this degree, there are a

few that deserve special mention. K. Kapheim and C. Beaudoin and J. Smith

helped in the collection of field data that were used in this dissertation. Jaime
Tanner and Heather Watts entered the program with me 6 years ago and have
been my close friends ever since. They have supported me in innumerable ways
since the very beginning and I can’t thank them enough for what they’ve done.
The importance of having a good office mate in one’s intellectual, professional
and personal development is often underestimated, but after spending my last
years at MSU sharing an office with Kevin Theis, I will not be making that
mistake. Despite his own, often intense workload, Kevin was always ready to
answer a question, debate a relevant or irrelevant topic, or simply join in a
voicing of frustration. His help, support and friendship throughout this process
have been critically important. Kevin, Jaime and Heather also provided valuable
comments and suggestions throughout all phases of my research and I am
indebted to them for their help.

Much of the data presented in this dissertation were not easy to collect,
and demanded long hours of radio-tracking at all hours of the day and night.
Without the help of Sofia Wahaj and Matt Gibbons, one of which was present
during almost all overnight hyena tracking sessions, these data would not have
been collected. Somehow, Sofi and Matt persevered through all-night follows,
dead car batteries in the middle of a dark grass plain, food deprivation, sleep
deprivation, and the general uneasiness that comes with driving off-road in tall
grass, in the dark. I am sure I have not thanked them enough for their part in
maintaining my general sanity and contentment in the field, and in making this

dissertation possible.

vi

Many of my ideas regarding this dissertation were born from the earlier
work of Erin Boydston, a previous member of the Holekamp Lab. Her excellent
dissertation research paved the way for many of my ideas and objectives and I
truly appreciate not only her previous efforts on the project, but also her
assistance with my own research efforts here. I received critical assistance in the
lab from undergraduates D. Katan, L. Kierepka and K. Lawracy as well as from
S. Congdon who all provided helpful assistance with certain analyses. I thank
Meredith Evans for helping with satellite image acquisition and analysis, Bilal Butt
for freely sharing books, spatial data, and advice on a wide-array of topics, and
Ashton Shortridge for taking the time to offer suggestions on spatial data
analyses.

I also thank the rest of all my friends l have made here in Lansing, for their
kindness and support, especially Harry S. Butterfield, Doug Bruggeman, Max
Fulkerson, Toshi Yoshida and Scott Kissman. I thank Stephanie Wolf, for putting
up with me during the stresses of completing this dissertation, and for doing so
much to keep a smile on my face ever since we met. My family, in particular my
parents, have always encouraged me to do what makes me happy. When that
brought me to Kenya, they may have regretted that encouragement, but they

supported me through it all. I can’t thank them enough.

vii

TABLE OF CONTENTS

LIST OF TABLES .................................................................................................. ix
LIST OF FIGURES .............................................................................................. xiv
INTRODUCTION ................................................................................................... 1
CHAPTER 1 .
DAILY PATTERNS OF ACTIVITY IN THE SPOTTED HYENA (CROCUTA
CROCUTA) .......................................................................................................... 11
Introduction ............................................................................................... 12
Methods .................................................................................................... 15
Results ...................................................................................................... 26
Discussion ................................................................................................. 41
CHAPTER 2
ECOLOGICAL AND ANTHROPOGENIC INFLUENCES ON SPACE USE
PATTERNS IN THE SPOTTED HYENA (CROCUTA CROCUTA) ...................... 51
Introduction ............................................................................................... 51
Methods .................................................................................................... 56
Results ...................................................................................................... 73
Discussion ................................................................................................. 98
CHAPTER 3
EFFECTS OF AN OPEN REFUSE PIT ON SPACE USE PATTERNS IN THE
SPOTTED HYENA (CROCUTA CROCUTA) ..................................................... 109
Introduction ............................................................................................. 1 10
Methods .................................................................................................. 112
Results .................................................................................................... 119
Discussion ............................................................................................... 123
CHAPTER 4

SPATIAL, TEMPORAL, AND PHYSICAL CHARACTERISTICS OF LIVESTOCK
DEPREDATIONS BY LARGE CARNIVORES ALONG A KENYAN RESERVE

BORDER ............................................................................................................ 128
Introduction ............................................................................................. 1 29
Methods ................................................ 132
Results .................................................................................................... 144
Discussion ............................................................................................... 163

APPENDIX ......................................................................................................... 174

LITERATURE CITED ......................................................................................... 199

viii

LIST OF TABLES

Table 1.1. Mean (SE) activity and movement rates for male (M) and female (F)
spotted hyenas in the Masai Mara National Reserve, Kenya. Time periods are
as follows: Evening: 1400—1900 h; Night1: 1900—2300 h; Night2: 2300—0300 h;
Night3: 0300—0700 h; Morning: 0700—1100 h ...................................................... 35

Table 1.2 Mean (SE) activity and movement rates for female spotted hyenas
from either the Talek West (TW) or Mara River (MR) clans within the Masai Mara
National Reserve, Kenya. Time periods as in Table 1.1 and daylight is 0700—
1800 h .................................................................................................................. 39

Table 2.1. Periods identified in the study period during which den location was
stable in the Mara River (MR) and Talek West (TKW) clans. Logistic regression
modeling of resource use was based on data collected only during these periods.
Levels of prey and livestock for each period are represented by the mean # of
ungulates/km of transect and the mean # of livestock animals counted per
census drive during each period, respectively ..................................................... 74

Table 2.2. Mean (95% CI) values of continuous variables recorded for hyena
tracking locations (“used”), and 5000 random locations (“available”) within the
territories of the Mara River (MR) and Talek West (TKW) clans. The first three
variables represent the straight-line distances (in meters) to the communal den
(Den), the nearest stream (Stream), nearest permanent stream (Perm stream)
and nearest vegetative cover (Cover). Livestock use values (LUV) and prey
values were based on grid surfaces created from livestock censuses and prey
sampling as described above. Only values that did NOT differ based on 95% Cl
are in bold ............................................................................................................ 80

Table 2.3. The percent of hyena radiolocations (“used”) and 5000 random
locations (“available") within the territories of the Mara River (MR) and Talek
West (TW) clans that were associated with each of three vegetation classes.
Periods indicated on the left represent time periods during which hyena locations
were collected (see Table 2.1 for period descriptions and location sample

sizes) .................................................................................................................... 82

Table 2.4. Logistic regression results for Mara River modeling periods based on
hyena locations collected during all times of day and night ................................ 86

Table 2.5. Logistic regression results for Talek West modeling periods based on
hyena locations collected during all times of day and night ................................. 87

Table 2.6. Modified odds ratios for models created based on all hyena tracking
locations (“All”: Table 2.4 and 2.5), and tracking locations from grazing hours
(GH: 0900—1800 h), and livestock-free hours (LFH: 1900—0800 h). Numbers

indicate the % change in probability of a location being identified as “used” by a
hyena, that is associated with a 100 m increase in distance to the den, nearest
stream, or vegetative cover, and a 5-unit increase in prey value. Variables not
selected for final models were assigned a value of 0. Not all selected variables
are shown ............................................................................................................ 89

Table 2.7. Pearson correlation coefficients between 3 ecological variables used
in logistic regression modeling of space use by members of the Mara River (MR)
and Talek West (TKW) clans. Variables listed are the distance to the communal
den location (Dist to Den), the livestock use intensity value (LUV) and the prey
value. Correlations are based on all locations used for modeling during each
period (Le. 5000 random locations plus all tracking locations for the period) ...... 93

Table 2.8. Monthly livestock use intensity values (LUVs) for radiolocations of
adult hyenas in the Talek West clan (95% Cl) collected during the entire month
(Overall Hyena), during only grazing hours (GH: 0900-1800 h) and livestock-free
hours (LFH: 1900-0800 h). For comparison, the average LUV for all cells in the
clan territory for each month (Territory) is also presented. Months with < 20
radiolocations in any time block were excluded. When the average LUV of hyena
locations is significantly greater than the territory LUV, “S” indicates selection for
livestock use areas relative to availability, and “A” indicates avoidance .............. 95

Table 2.9. Average livestock use intensity values (LUVs) for radiolocations of
adult hyenas in the Talek West clan (95% CI) collected during grazing hours
(GH: 0900—1800 h) and during livestock-free hours (LFH: 1900—0800 h). LUVs
were derived from fixed-kernel grid surfaces based on all herds observed during
the monitoring period of each hyena. Hyenas monitored for < 18 months were
excluded. For comparison, the average LUV for all cells in the Reserve portion of
the clan territory, based on entire study period, was 12.6 (18.6). There was no
overall difference in average LUV of locations from OH and LFH using individual
hyena means as samples (paired t= -1.22; p = 0.25) ......................................... 96

Table 2.10. Monthly average distances to, and % inside, the livestock core area
(LCA) for hyena radiolocations collected during grazing hours (GH: 0900—1800 h)
and livestock-free hours (LFH: 1900—0800 h). The LCA was the area of most
intense livestock use during the study period. Months with < 20 radiolocations in
any time block were excluded, except for the last row ............................................. 97

Table 3.1. Average distances (SE) of adult female hyena tracking locations to
the refuse pit for the 10 months prior to the closing of the pit (pre-closure) and the
19-months following the closure (post-closure). The rank group of each hyena is
indicated after each name (H=High, L=Low) ..................................................... 121

Table 4.1. Univariate comparisons between mean values i standard error of
nine independent variables recorded at attacked and unattacked Maasai bomas
(villages) within 2 km of the northeastern border of the Masai Mara National

Reserve, Kenya. P-values are based on Mann-Whitney U tests. Variables
indicated by asterisks are intercorrelated .......................................................... 155

Table 4.2. Univariate comparisons between mean values i standard error of
seven independent variables recorded at attacked and unattacked sheep/goat
enclosures within 2 km of the northeastern border of the Masai Mara National
Reserve, Kenya. P-values are based on Mann-Whitney U tests for continuous
variables and Fisher exact tests for proportions. None of the variables examined
here were intercorrelated .................................................................................. .157

Table 4.3. Logistic regression models predicting the probability that a Maasai
boma (village) will be attacked by large carnivores. Presented models were
selected based on the minimum AIC value of all model combinations using 6
uncorrelated continuous variables ..................................................................... 160

Table 4.4. Logistic regression models predicting the probability that a Maasai
sheep/goat (shoat) enclosure will be attacked by large carnivores. Presented
models were selected based on the minimum AIC value of all model
combinations using five uncorrelated continuous variables and one categorical
variable .............................................................................................................. 162

Table A1 Logistic regression results showing all models within 2.0 AIC points of
the lowest AIC model for all Mara River modeling periods based on hyena
locations collected during all times of day and night. Final selected models are in
bold. Only one model was identified for MR2 and MR3 ..................................... 174

Table A2. Logistic regression results showing all models within 2.0 AIC points of
the lowest AIC model for the first Talek West modeling period (TKW1) based on
hyena locations collected during all times of day and night. The final selected
model is in bold .................................................................................................. 175

Table A3. Logistic regression results showing all models within 2.0 AIC points of
the lowest AIC model for the second Talek West modeling period (T KW2) based
on hyena locations collected during all times of day and night. The final selected
model is in bold .................................................................................................. 176

Table A4. Logistic regression results showing all models within 2.0 AIC points of
the lowest AIC model for the third Talek West modeling period (TKW3) based on
hyena locations collected during all times of day and night. The final selected

model is in bold .................................................................................................. 177

Table A5. Logistic regression results showing all models within 2.0 AIC points of
the lowest AIC model for the fourth Talek West modeling period (TKW4) based
on hyena locations collected during all times of day and night. The final selected
model is in bold .................................................................................................. 178

xi

Table A6. Logistic regression results showing all models within 2.0 AIC points of
the lowest AIC model for the first Mara River modeling period (MR1) based on

hyena locations collected during grazing hours (0900—1800 h). The final selected
model is in bold .................................................................................................. 179

Table A7. Logistic regression results showing all models within 2.0 AIC points of
the lowest AIC model for the first Mara River modeling period (MR1) based on
hyena locations collected during livestock-free hours (1900—0800 h). The final
selected model is in bold .................................................................................... 180

Table A8. Logistic regression results showing all models within 2.0 AIC points of
the lowest AIC model for the second Mara River modeling period (MR2) based
on hyena locations collected during grazing hours (0900—1800 h). The final
selected model is in bold .................................................................................... 181

Table A9. Logistic regression results showing all models within 2.0 AIC points of
the lowest AIC model for the second Mara River modeling period (MR2) based

on hyena locations collected during livestock-free hours (1900—0800 h). The final
selected model is in bold .................................................................................... 182

Table A.10. Logistic regression results showing all models within 2.0 AIC points
of the lowest AIC model, as well as the final selected model (in bold), for the third
Mara River modeling period (MR3). Models are based on hyena locations
collected during grazing hours (0900—1800 h) ................................................... 183

Table A.11. Logistic regression results showing all models within 2.0 AIC points
of the lowest AIC model for the third Mara River modeling period (MR3) based on
hyena locations collected during livestock-free hours (1900—0800 h). Only one
model was identified .......................................................................................... 184

Table A.12. Logistic regression results showing all models within 2.0 AIC points
of the lowest AIC model, as well as the final selected model (in bold), for the first
Talek West modeling period (TKW1). Models are based on hyena locations
collected during grazing hours (0900—1800 h) ................................................... 185

Table A.13. Logistic regression results showing all models within 2.0 AIC points
of the lowest AIC model, as well as the final selected model (in bold), for the first
Talek West modeling period (TKW1). Models are based on hyena locations
collected during livestock-free hours (1900-0800 h) ......................................... 186

Table A.14. Logistic regression results showing all models within 2.0 AIC points
of the lowest AIC model, as well as the final selected model (in bold), for the
second Talek West modeling period (TKW2). Models are based on hyena
locations collected during grazing hours (0900—1800 h) ................................... 188

xii

Table A.15. Logistic regression results showing all models within 2.0 AIC points
of the lowest AIC model for the second Talek West modeling period (TKW2)
based on hyena locations collected during livestock-free hours (1900—0800 h).
The final selected model is in bold .................................................................... 190

Table A.16. Logistic regression results showing all models within 2.0 AIC points
of the lowest AIC model for the third Talek West modeling period (TKW3) based
on hyena locations collected during grazing hours (0900—1800 h). The final
selected model is in bold .................................................................................... 191

Table A.17. Logistic regression results showing all models within 2.0 AIC points
of the lowest AIC model for the third Talek West modeling period (TKW3) based
on hyena locations collected during livestock-free hours (1900—0800 h). The final
selected model is in bold .................................................................................... 192

Table A.18. Logistic regression results showing all models within 2.0 AIC points
of the lowest AIC model for the fourth Talek West modeling period (TKW4).
Using only hyena locations collected during grazing hours (0900-1800 h), only
one model was identified ................................................................................... 193

Table A.19. Logistic regression results showing all models within 2.0 AIC points

of the lowest AIC model for the fourth Talek West modeling period (TKW4) based
on hyena locations collected during livestock-free hours (1900—0800 h). The final
selected model is in bold .................................................................................... 194

Table A.20. Results from all logistic regression models considered for the
prediction of the probability of hyena attack on livestock within Maasai bomas
(villages). Results shown are from all models within 2.0 AlCc points of the
optimal model (considered Model #1, also shown in Table 4.3) ........................ 195

Table A.21. Results from all logistic regression models considered for the
prediction of the probability of leopard attack on livestock within Maasai bomas
(villages). Results shown are from all models within 2.0 AlCc points of the
optimal model (considered Model #1, also shown in Table 4.3) ........................ 196

Table A.22. Results from all logistic regression models considered for the
prediction of the probability of hyena attack on livestock within sheep/goat (shoat)
enclosures. Results shown are from all models within 2.0 AlCc points of the
optimal model (considered Model #1, also shown in Table 4.4) ........................ 197

Table A.23. Results from all logistic regression models considered for the
prediction of the probability of leopard attack on livestock within sheep/goat
(shoat) enclosures. Results shown are from all models within 2.0 AlCc points of
the optimal model (considered Model #1, also shown in Table 4.4) .................. 198

xiii

LIST OF FIGURES

Figure 1.1. Territory boundaries (dotted lines) of the three monitored spotted
hyena social groups (clans) in the Masai Mara National Reserve, Kenya.
Triangles indicate the 4 main tourist lodges in the study area, open circles
represent Maasai villages, and a star indicates the town of Talek. Prey sampling
transects are indicated with checkered lines ....................................................... 17

Figure 1.2. Activity pattern and movement rates of radiocollared adult spotted
hyenas in the Masai Mara National Reserve, Kenya. Data are based on direct
observations during long-term (2—1 5 h) follows of males and females. Number of
individuals sampled varies per hour (n = 15—19). Plotted values for each hour
(eg. 1200 h) represent activity recorded in the following hour block (Le. 1200—
1259 h). The black bar extends over hour blocks characterized by darkness
throughout the year .............................................................................................. 27

Figure 1.3. Bouts of activity (solid black bars) during 14 composite 24-h follow
cycles for 13 individual radiocollared spotted hyenas. Active bouts were >5
consecutive active minutes bounded on both ends by 5 consecutive minutes of
inactivity. Pictured 24-h cycles were those for which the night was described by
no more than 2 follow segments. The letter “a" indicates a pair of follow cycles
conducted on the same individual. Start and end points of each follow segment
are indicated by a “v”. For example, the topmost follow cycle is composed of
three nonconsecutive follow segments conducted from 1730-0100 h, 0100-1000
h, and 1000-1730 h. Asterisks indicate hunts and fresh kills on which the focal
animal fed. Hours of darkness throughout the year fall within the shaded

area ...................................................................................................................... 30

Figure 1.4. Mean number of group size changes observed and mean % time
spent socializing during follows of radiocollared spotted hyenas. Data are from
direct observations of 16 different individuals (11 female, 5 male) during 22
composite follows, each covering an entire 24-h period in nonconsecutive
segments. Other notation is as in Figure 1.2 ...................................................... 31

Figure 1.5. Mean % time spent engaged in various behaviors for 5 male and 11
female spotted hyenas during follows covering the entire 24—h period in
nonconsecutive segments. Miscellaneous activity included standing, and sniffing
objects or the ground. Traveling included both directed movement and wandering
behavior. Significant differences (Mann—Whitney U-test, P < 0.05) are indicated
by an asterisk ....................................................................................................... 33

Figure 1.6. A) Sex differences in mean % time spent active by radiocollared

adult spotted hyenas. Data are based on direct observation during long-term (2-
15 h) follows of males and females. Number of individuals sampled varies per

xiv

hour (females n = 9-11; males n = 5-8). Figure notation is as in Figure 1.2. B)
Mean hourly movement rates of radiocollared adult spotted hyenas excluding
data collected during hour blocks when focal females nursed their cubs. Number
of individuals sampled varies per hour (females n = 7-11; males n = 5—7) ......... 34

Figure 1.7. The mean % time spent A) active and B) socializing by radiocollared
adult female spotted hyenas from two different social groups (clans). Data are
based on direct observation during long-term (2—15 h) follows. Number of
individuals sampled varies per hour (Talek West n = 5—6; Mara River n = 4—5 for
activity and n = 3—5 for socializing). Figure notation is as in Figure 1.2 ............... 40

Figure 1.8. Percent of A) evening and B) morning scans at active communal
dens in which no hyenas were present during each of 5 half-hour blocks for the
Mara River Clan and the Talek East and West Clans .......................................... 42

Figure 2.1. The location of the territory boundaries of the two study clans.
Ungulate transects are shown with barred lines and Maasai villages within 2 km
of the Reserve are shown with open circles. Note the extension of the Talek
West territory outside the Reserve boundary ....................................................... 58

Figure 2.2. Vegetation class maps for the territories of the Mara River and Talek
West hyena clans. Permanent and seasonal streams are shown in blue, and the
three vegetation classes are grass plain (yellow), shrubland (brown) and riparian
forest (green) ....................................................................................................... 66

Figure 2.3. Prey distribution grid surfaces (cell length = 500 m) for each of the 3
modeling periods for the Mara River clan. Darker colors indicate higher estimated
prey densities. The communal den location used during each period, and the
clan territory boundary, is indicated on each map. Note that prey distributions
indicated outside territory boundaries are not accurate ....................................... 76

Figure 2.4. Prey distribution grid surfaces (cell length = 500 rn) for each of the 4
modeling periods for the Talek West clan. Darker colors indicate higher
estimated prey densities. The communal den location used during each period,
and the clan territory boundary is indicated on each map. Note that prey
distributions indicated outside territory boundaries are not accurate ................... 77

Figure 2.5. Livestock utility distribution grid surfaces for each of the 4 modeling
periods in the Talek West clan. Darker colors indicate higher intensity of use by
livestock. The communal den location used during each period, and the clan
territory boundary is indicated on each map ......................................................... 78

Figure 2.6. The mean % difference between values of continuous variables
measured at hyena locations (“used”) and those measured at 5000 random
locations (“available”) within the respective clan territories. A % difference was
calculated for each of 3 modeling periods in Mara River (MR), and 4 periods in

XV

Talek West (TKW); the mean of these values is indicated above. The first 4
variables represent distances; therefore negative % deviations indicate hyena
selection for proximity to these features. Variable descriptions, the data from
each individual model, and the associated statistical comparisons are shown in
Table 2.2 .............................................................................................................. 79

Figure 2.7. The mean % difference between the proportion of hyena locations
and the proportion of 5000 random locations (“available”) within the respective
clan territories that were located in each of 3 vegetation class. The % difference
was calculated for each of 3 modeling periods in Mara River (MR), and 4 periods
in Talek West (TKW); the mean of these values is indicated above. The data
from each individual model and associated statistical comparisons are shown in
Table 2.3 .............................................................................................................. 81

Figure 2.8. Locations at which adult male and female Mara River hyenas were
radiotracked within their territory during 3 modeling periods (see Table 2.1).
Locations at the communal den for females with den-dwelling cubs were
excluded. Permanent and seasonal streams are indicated along with the
communal den (orange triangle) used during each period ................................. .83

Figure 2.9. Locations at which adult male and female Talek West hyenas were
radiotracked within their territory and within the Reserve during 4 modeling
periods (see Table 2.1). Locations at the communal den for females with den-
dwelling cubs were excluded. Permanent and seasonal streams are indicated
along with the communal den (orange triangle) used during each period ........... 84

Figure 3.1. Territory of the Mara River hyena clan and its location within the
Masai Mara National Reserve. Hatched lines represent ungulate prey sampling
transects. The refuse pit (filled circle) and tourist lodge are also indicated ...... 115

Figure 3.2. The 95% home ranges (solid dark lines) and 50% core areas (dotted
lines) based on locations from 5 radiocollared adult female hyenas collected
during the 10 months before (A) and 19 months after (B) the closure of the refuse
pit. Pit location is indicated as in Fig. 1, and the locations of all communal dens
utilized during the respective periods are indicated by solid triangles ............... 122

Figure 3.3. Mean number of high-ranking (HR) females, low—ranking (LR)
females, and immigrant male hyenas (5 individuals per group) that were seen at
18 refuse pit observation sessions. Sessions included were those at which at
least one hyena was observed. Whiskers indicate 95% confidence

intervals .............................................................................................................. 124

Figure 4.1. Locations of Maasai bomas (villages) of the Talek (filled circles),
Ntipilikwani (open circles), and Olosogon (triangles) locales from which
information on livestock depredation by large carnivores was collected from
March 2003 to April 2004. Only bomas that contained livestock and were located

xvi

within 2km of the Reserve border are included. Dashed lines indicate prey
transects and the town of Talek is indicated with a star ..................................... 133

Figure 4.2. Configuration of a typical Maasai village (boma). Cattle from all
household heads are housed together in the shared central corral. Individual
homesteads maintain separate enclosures for their own small stock. Figure
adapted with permission from Spencer (2003) pg 45 ........................................ 136

Figure 4.3. The percent of the total losses of sheep and goats, cattle, and all
livestock species that were attributed to spotted hyena, lion, and leopard
depredation. Percentages shown are out of a total of 147 livestock killed during a
14-month period along the northeast border of the Masai Mara National Reserve.
Kenya ................................................................................................................. 145

Figure 4.4. Involvement of lion, leopard and hyena in 4 categories of livestock
depredation incidents along the northeast border of the Masai Mara National
Reserve during a 14-month study period. Total numbers of attacks of each type
are listed in parentheses after category headings. "B" indicates attacks at the
boma; "G" indicates attacks while grazing ......................................................... 147

Figure 4.5. Total monthly livestock attacks by predators recorded within 2 km of

the northeastern border of the Masai Mara National Reserve, Kenya, from March
2003 to April 2004. Conflicts included attacks by lions, leopards, and hyenas that
resulted in either death or injury of livestock ...................................................... 150

Figure 4.6. Correlation between monthly rainfall (mm) and total predator incident
reports collected each month during a 14-month study period .......................... 151

Figure 4.7. Mean percent of individual hyena home ranges (95% fixed-kernel)
that were located outside the Masai Mara National Reserve based on locations
collected during months of low, medium and high frequencies of livestock
depredation events involving hyenas from March 2003-April 2004. Whiskers
indicate 95% confidence intervals ...................................................................... 153

Figure 4.8. Relative attack rates by leopard and hyena on the two most common
types of sheep/goat enclosures within 2 km of the northeastern Reserve border.

Significant differences (Fisher exact test for comparison of proportions; p < 0.05)
are indicated by an asterisk ............................................................................... 158

Images in this dissertation are presented in color.

xvii

INTRODUCTION

Worldwide, habitat suitable for wildlife is becoming increasingly rare due to
human population growth and increasing rates of urbanization and habitat
conversion. Because most mammalian carnivores exist at the top of their local
food webs, they tend to range far more widely in search of food and mates than
do other terrestrial mammals. As a result, carnivores may be more likely to come
into contact with human populations than other species (Woodroffe 2001).
Because large carnivores can rarely survive in human-altered habitats, they are
particularly sensitive to habitat fragmentation (Woodroffe & Ginsberg 1998,
2000), and their low reproductive output and long-distance dispersal make them
highly vulnerable to extinction (Sunquist & Sunquist 2001). However, their
continued presence is often critical for proper ecosystem function (Berger 1999;
Crooks & Soulé 1999; Ripple et al. 2001; Terborgh 2001).

Unfortunately, because local people often perceive predators as a direct
threat to themselves or their livestock, sustained coexistence between humans
and large carnivores is rarely possible outside of protected areas (Woodroffe &
Ginsberg 2000). Therefore, as habitat is settled and developed, protected areas
become the last remnants of suitable predator habitat. Today, however, even
within these areas many carnivore populations continue to be affected by
anthropogenic disturbance. Throughout the world, humans cause most adult
mortality among large carnivores, including those inhabiting protected areas

(Woodroffe & Ginsberg 2000). Not surprisingly, most adult mortality occurs when

carnivores range beyond reserve borders. Thus, while habitat loss and
persecution are restricting large carnivores to protected areas, 'edge effects'
caused by human activity along reserve borders continue to reduce the viability
of populations inside protected areas. As a result, research that investigates the
effects of human activity on carnivores in and around protected areas, as well as
the characteristics of human-carnivore interactions in these systems, becomes
increasingly critical to the development and maintenance of carnivore
conservation efforts around the world.

Despite the socio-political and economic complexities of carnivore
conservation efforts, it is widely recognized that knowledge of the behavioral
ecology of species of concern can be critical to successful management and
conservation decisions (Caro 1999a). For example, detailed knowledge of
behavior relating to a species’ adaptability, flexibility, risk taking behavior,
spacing, mating system, territoriality and habitat use are all directly related to
their vulnerability to extinction and their resilience to disturbance or change
(Arcese et al. 1997; Brashares 2003; Caro & Durant 1995; Weaver et al. 1996).
Furthermore, behavioral data are critical to effective management of human-
wildlife conflicts, as they inform the design of conflict prevention measures and
identify the vulnerability of subsets of populations to conflict involvement (Arcese
et al. 1997; Treves & Karanth 2003).

The Masai Mara National Reserve (hereafter the Reserve) in southwest
Kenya offers an ideal setting in which to study human-carnivore interactions.

The situation developing in the Reserve and surrounding areas typifies that

facing numerous large carnivore species worldwide. Growth of local human
populations in areas far north of the Reserve has sparked the clearing of
savanna and woodlands to provide room for increased cultivation and settlement,
constricting the habitat available to wildlife (Broten & Said 1995). Expanding
cultivation, primarily in the form of large-scale wheat farming, has spread south
toward the Reserve (Homewood et al. 2001) and has had the additional effect of
limiting grazing land available to Maasai pastoralists (Serneels et al. 2001).
Buffer zones surrounding the Reserve, which have traditionally been subjected to
low intensity pastoralist or agropastoralist land use, are therefore becoming
increasingly degraded (Serneels et al. 2001). Although livestock numbers have
remained relatively stable in this region (Lamprey & Reid 2004; Serneels &
Lambin 2001), livestock grazing inside the Reserve, though prohibited, has
increased markedly in recent years and was described by the Kenya Wildlife
Service as a ‘rampant’ problem in 1997 (Muriuki & Mulama 1997). Although
herds are occasionally seen deep in the Reserve, the vast majority of trespassing
occurs along the northeastern border (Muriuki et al. 2000). The human
population itself is increasing along the Reserve border. All the lands bordering
the Reserve to the north are communally owned, grazed and managed as “group
ranches”. On the Koyake group ranch, one of the largest group ranches
bordering the northeastern portion of the Reserve, the human population is
doubling every 15 years (Lamprey & Reid 2004), providing an increasing threat

of direct persecution for carnivores ranging outside Reserve boundaries.

Although the spotted hyena (Crocuta crocuta), like most large African
carnivores, has suffered a reduction in the extent of its historical range (Mills &
Hofer 1998), it is recognized by the IUCN as a species of Lower Risk (IUCN
2006), and remains the most abundant and widespread large carnivore in Africa.
The hyena's catholic diet has contributed to its ability to survive in diverse
environments characterized by variable food resources, habitat characteristics,
and competitors. Therefore, the responses ofthis species to human disturbances
should represent conservative indicators of how other top predators, including
those that are threatened and endangered, are likely to respond to similar
perturbations. lf human activity has negative effects on this highly adaptable and
resourceful species, then the future of less resilient carnivore species in the
same ecosystems is brought into serious question.

An intensive field study of the behavioral ecology of spotted hyenas in the
Reserve began in 1988, under the direction of Dr. Laura Smale and Dr. Kay
Holekamp, and continues to this day. The study has focused on members of a
single hyena social group known as the Talek clan (which has since split into two
clans; Talek West and Talek East) that defends a territory along the northeastern
border of the Reserve. Due to its location, this clan has been exposed to
increasing levels of human activity over the years of the study and currently
interacts regularly with Maasai pastoralists and their livestock. The existence of
this long-term study provided a unique opportunity to investigate the details of

human-carnivore interactions in and around a protected reserve, and to

incorporate detailed knowledge about the behavior and social structure of this
species and about the individual identity of study animals.

The overarching objective of this dissertation research was to provide
quantitative information on the effects of anthropogenic disturbances on spotted
hyena movements and activity. However, whereas hyenas and other carnivores
must adapt to the presence of humans in this and many other systems, so must
humans adapt to the presence of carnivores. The Maasai people of Kenya that
reside around the Reserve have a culture rich in pastoral tradition. Livestock
play a central role in all aspects of their lives, and therefore conflict with large
carnivores is an additional and inevitable component of their culture. Because
the majority of carnivore mortality in protected areas is a result of human
persecution, largely occurring outside reserve boundaries (Woodroffe & Ginsberg
2000), human-carnivore conflict must be mitigated, and tolerance of carnivores in
human landscapes increased, to ensure the long-term persistence of many large
carnivore populations. With this perspective, I focused one portion of my
research on an investigation of carnivore-livestock conflict along the Reserve
border. It is my hope that the results of this dissertation will be useful in guiding
future management efforts in regard to spotted hyenas, as well as other large
carnivores, and will elucidate the relationship between human activities and the
carnivore populations struggling to exist in human-dominated landscapes. I also
suggest that behavioral changes in animal populations subjected to varying
degrees of human disturbance may be used as early indicators of potentially

more severe consequences for population demography and viability.

I begin this dissertation, in Chapter 1, with an investigation into the activity
patterns of spotted hyenas. Members of the Talek West clan have been
monitored and closely observed since the start of this study in 1988. Because
the clan territory is located along the heavily populated and intensely grazed
northeastern border of the Reserve, this clan is subjected to frequent interactions
with humans. Previous research suggested that hyenas in this clan have
become more nocturnal over time, concurrent with an increase in the intensity of
livestock grazing to which they were exposed (Boydston et al. 2003b). Similar
modifications to activity patterns have been seen in numerous carnivore species
resulting from either specific disturbances such as hunting or recreational
activities (Kitchen et al. 2000; Olson et al. 1998), or from high overall levels of
human activity and urbanization within territories (Ciucci et al. 1997; Lucherini et
al. 1995; McClennen et al. 2001; Riley et al. 2003).

To address the question of whether Reserve hyenas exhibit similar activity
pattern shifts in response to livestock grazing, l utilize a unique comparative
approach, ideal for identifying the effects of human disturbance in natural settings
(Arcese 8 Sinclair 1997; Caro 1999a). In 2001, I began studying hyenas in a
clan located in the center of the Reserve, the Mara River clan. The territory
defended by this clan is similar in all ecological aspects, including habitat and
prey resources, to that of the Talek West clan, and it is located less than 6 km
from the western edge of Talek West territory. However, due to its distance from
the Reserve boundary, no livestock grazing occurs in the Mara River clan

territory. Throughout this study this clan serves as a critical baseline control

group and allows formal tests of predictions derived from hypotheses suggesting
the effects of human activity on carnivore behavior. Specifically in this first
chapter, I hypothesized that intense livestock grazing, which occurs during the
day, has influenced the activity patterns of Talek West hyenas. I therefore
predicted that differences in activity would be apparent between the Talek West
and Mara River clans. In addition to providing detailed descriptive data on
spotted hyena activity, including an investigation of the influence of sex and rank
on activity, my results document clear differences in activity patterns between the
two clans that correlate closely with the activity of grazing livestock. This chapter
is currently in press at the Journal of Mammalogy, therefore references to it in
Chapters 2—4 cite this manuscript.

Chapter 2 is a comprehensive investigation into the ecological
determinants of space use in the spotted hyena and represents a logical
continuation of the previous work of Boydston (2001) in the same Reserve. Here
again I used a comparative approach to investigate the influence of livestock
grazing on space use decisions by individual clan members. Although studies
abound of space use by carnivores, few are able to associate space use with
both landscape features and more variable ecological features such as prey
distribution. Even fewer studies have investigated the influence of livestock
grazing on the space use decisions of large carnivores and none of these have
been conducted in East Africa. This chapter also focuses on the modifications in
space use that may be available to spotted hyenas to minimize costs associated

with livestock-related disturbance. Despite predictions to the contrary, data

presented here indicate that Talek West hyenas did not actively avoid areas used
by livestock, nor did they shift their temporal use of space to reduce use of
heavily grazed areas during times of day when grazing herds were present.
However, logistic regression modeling in both the Mara River and Talek West
clan showed that the presence of vegetative cover, which has been shown to be
highly variable over time in this ecosystem (Dublin 1995; Serneels et al. 2001),
may be critically important to the successful coexistence of pastoralists and
hyenas.

Although sheltered from most human activity, the hyenas in the Mara
River clan were exposed for some time to the presence of an easily accessible,
anthropogenic, non-livestock food source in form of a refuse pit near the border
of their territory. In Chapter 3, I investigate the influence of this food source on
the movement patterns of Mara River hyenas. I show not only that this small
food source influenced movements of clan hyenas, but also that frequency of use
of this food source was dependent on the local abundance of prey, and varied
with an individual’s social rank. These results are directly relevant to livestock
depredation issues, as human refuse is often readily available near villages, and
its presence has been suggested to influence rates of carnivore-human conflict
(Beckmann & Berger 2003b) and livestock depredation rates (Mills & Hofer
1998). In addition, these findings indicate that behavioral variation among
population subgroups may result in individual variation in susceptibility to edge

effects. This chapter is currently under review at the African Joumal of Ecology.

The dissertation concludes in Chapter 4 with a detailed investigation of the
direct interactions between carnivores and Maasai pastoralists. There had been
sporadic reports of livestock lost to hyenas, as well as leopards and lions,
throughout the study since its inception (K. E. Holekamp, pers. comm).
However, the extent of these losses, the predators involved, and the factors
influencing rates of livestock loss were all unknown. Here I document the extent
of this depredation problem, and the relative involvement of each of the different
predators present in the region. Most importantly, I show that livestock
depredation has a clear and predictable association with rainfall, and that the rate
of depredation at nighttime livestock corrals by leopards and hyenas is
associated with fence type as well as by village location in the landscape. These
findings and others result in clear recommendations for conflict mitigation that
can be applied not only to this region but also across East Africa. Furthermore,
the relationship of depredation rates to a combination of ecological and
husbandry-related variables is supported by other recent work (Mech et al. 2000;
Michalski et al. 2006; Ogada et al. 2003; Patterson et al. 2004); therefore, the
implications of these results can potentially be applied to depredation
management worldwide. This chapter was published in 2006 in Biological
Conservation, therefore references to this chapter in earlier chapters cite this
manuscript.

Because all Chapters presented here were prepared in manuscript form,
and because the field work associated with these chapters, and the final

preparation of the manuscripts was a truly collaborative effort, I have used the

term “we” instead of “l” throughout the remainder of this dissertation, as l have

done in the manuscripts themselves.

10

CHAPTER ONE

Kolowski, J.M., D. Katan, K. R. Theis, and K. E. Holekamp. 2007. Daily patterns
of activity in the spotted hyena. Journal of Mammalogy (August issue).

11

CHAPTER ONE
DAILY PATTERNS OF ACTIVITY IN THE SPOTTED HYENA
INTRODUCTION

As in other mammals, the daily activity patterns of terrestrial carnivores
result from both endogenous biological rhythms and behavioural adaptations to a
changing environment (Daan 1981; Rusak 1981). Optimal patterns of activity
reflect the influence of daily temperature variation (Avenant & Nel 1998;
Garshelis 8. Pelton 1980), predation risk (Drew & Bissonette 1997; Geffen &
Macdonald 1993; Lima & DIN 1990) and prey availability (Ferguson et al. 1988;
Garshelis & Pelton 1980; Lariviere et al. 1994; Lode 1995). However, individual
characteristics such as sex and reproductive condition may also influence
patterns of daily activity (Daan & Aschoff 1982; Lariviere & Messier 1997; Paragi
et al. 1994; Zalewski 2001). Furthermore, in gregarious carnivores whose
societies are structured by linear dominance hierarchies, social status may also
contribute to variation in activity patterns. Given increased human presence on
landscapes worldwide, large carnivores have been shown to alter their natural
patterns of activity to avoid human disturbances ranging from tourist or
recreational activity (Machutchon et al. 1998; Olson et al. 1998) to direct
exploitation by hunting (Andelt 1985; Kitchen et al. 2000). Therefore, where
present, humans are likely to further influence observed activity patterns of free-
living carnivores. Here we examine effects of these variables on activity patterns

observed among spotted hyenas (Crocuta crocuta).

12

All four extant species in the family Hyaenidae are reported to be primarily
nocturnal (Bothma 8 Nel 1980; Kruuk 1976; Mills 1984, 1990). However, the
activity pattern of the spotted hyena, a gregarious carnivore found throughout
sub-Saharan Africa, appears to be the most flexible, with activity often extending
into periods of daylight (Kruuk 1972), or even occurring at mid-day (Rainy &
Rainy 1989). The only systematic data describing Crocuta’s activity pattern
come from the Serengeti ecosystem in Tanzania (Kruuk 1972) and from South
Africa (Henschel 1986; Mills 1990). In both locations, hyenas exhibited a
strongly nocturnal activity pattern with somewhat reduced activity during the
middle of the night (Kruuk 1972; Mills 1990). However, within the Serengeti, the
timing of evening activity peaks varied unpredictably over the course of months
or weeks (Kruuk 1972). Furthermore, overall levels of hyena activity and
movement rates were approximately twice as high in the Kalahari as in the
Serengeti (Kruuk 1972; Mills 1990). Thus, patterns of activity are clearly variable
in this species, yet sources of this variation remain unclear. Although Henschel
(1986) documented sex differences in Crocuta activity and movement, the
influence of social rank on these variables has yet to be investigated.

Our long-term study of spotted hyenas provides an opportunity to
document in detail the activity patterns and movement rates of individual male
and female Crocuta of various social ranks, and to investigate the influence of
human presence on hyena activity. Livestock grazing has already been shown to
influence patterns of space utilization by the Crocuta in our study population, and

additional effects of grazing on hyena activity patterns were suggested by those

13

data (Boydston et al. 2003b); however, the effect of daytime cattle grazing on
carnivore activity patterns has not been investigated in any study system.

Our goals here were to (1) describe the general activity pattern and time
budget of free-living Crocuta in Kenya, (2) investigate the influence of sex and
social rank on activity patterns and time budgets and (3) investigate the effects of
daytime human disturbance on hyena activity patterns and timing of den use.
Because reproductive success among male hyenas is associated with their
frequency of social interaction with the group's breeding females (East et al.
2003; Szykman et al. 2001), we expected that male movement and activity rates
would exceed those of females, due to a need to regularly interact with as many
females as possible. Because hyenas of low social rank have low priority of
access to food resources, we expected they would exhibit higher activity and
movement rates than hyenas of high rank. Finally, we were able to assess the
influence of human disturbance on hyena activity patterns by comparing activity
between members of neighboring social groups that differed in their exposure to
both livestock grazing and tourist visitation during daylight hours. If hyena
activity is affected by human presence on the landscape, then we expected
activity rhythms to differ between these groups; specifically, hyena activity and
den use in disturbed areas should be reduced when humans and cattle are

present.

14

METHODS

Study area

Our study was conducted from 2001—2004 in and around the Masai Mara
National Reserve (1500 kmz, hereafter the Reserve) in southwestern Kenya
(1 °40'S, 35°50’E). The Reserve consists primarily of rolling grassland and
scattered bushland (predominantly Croton and Euclea species), with riparian
forest along the major watercourses. Average annual rainfall in the study area
from 2001—2004 was 1305 mm. Most rainfall occurs during two wet seasons: the
“short rains” in November—December, and the “long rains” in March—May. Mean
monthly daytime temperatures in 2003 averaged 283°C (range: 25.1—32.3°C),
with the lowest temperatures occurring from May to July. Monthly nighttime
temperatures averaged 138°C (range: 12.2—14.9°C). Because of the proximity
of our study site to the equator, sunset and sunrise times varied little throughout

the year with sunrise occurring from 0618—0648 h and sunset from 1828—1858 h.

Study populations

Spotted hyenas live in social groups called clans, and clan members
cooperatively defend a stable group territory. Each clan contains one to several
matrilines of adult females and their offspring, as well as one to several adult
immigrant males. Clans are rigidly structured by hierarchical rank relationships
(Frank 1986b; Kruuk 1972; Tilson & Hamilton 1984), with adult females socially
dominant to immigrant adult males (Kruuk 1972; Smale et al. 1993). Subadult

individuals of both sexes maintain their maternal ranks as long as they remain in

15

the natal clan (Smale et al. 1993). Although females are generally philopatric
(Frank 1986b), almost all natal males disperse between the ages of 2 and 5
years (East & Hofer 2001; Henschel & Skinner 1987; Van Horn et al. 2003).
Crocuta clans are fission-fusion societies in which individuals travel, rest and
forage in subgroups that typically change in composition many times each day
(Holekamp et al. 1997a; Kruuk 1972). Female Crocuta bear 1—2 (rarely 3) young
In isolated natal dens (Holekamp et al. 1996). Cubs are typically transferred to a
communal den at 2—5 weeks of age where they reside for the next 7-8 months
(Kruuk 1972). The communal den represents the social center of each clan’s
territory and most clan members visit it regularly. Den-independent cubs
generally continue to nurse from their mothers until they are approximately 11-14
months old (Holekamp et al. 1996).

We monitored adult (>3 years old) members of three different clans. The
adjacent territories of the Talek East (19.0 kmz, 29-35 members) and Talek West
(28.4 kmz, 47—55 members) clans were both located mainly within Reserve
boundaries but partially extended outside the northeastern border of the Reserve
into lands occupied by Maasai pastoralists (Figure 1.1). The dominant land uses
on Maasai-owned land were subsistence pastoralism and wildlife tourism. The
Talek region supports the highest density of Maasai settlements along the entire
northern border of the Reserve (Reid et al. 2003). Settlements within 2 km of this
northeastern Reserve border supported roughly 12,000 cattle and 16,500 sheep
and goats (Kolowski & Holekamp 2006), some of which were grazed illegally

within the Reserve. Hyenas from both Talek clans have been killed at villages

16

 

    
  

  
 

Masai Mara
Natl onal Reserve

     
 

Masai Mara
National Reserve

 
 
 

 

 

 

MaaRiver '.
Terrritor 3.» 3;“
. J g o.
\5' 03°
k , A; ,2) g, a. mm
.1 .0. i. 5. 0 g a. = ------- x
4. w w..., j \ . - 6% -

Territory 3
'.

Figure 1.1. Territory boundaries (dotted lines) of the three monitored spotted
hyena social groups (clans) in the Masai Mara National Reserve, Kenya.
Triangles indicate the 4 main tourist lodges in the study area, open circles
represent Maasai villages, and a star indicates the town of Talek. Prey sampling

transects are indicated with checkered lines.

near the Reserve during livestock depredation attempts (Kolowski & Holekamp
2006). The areas outside the Reserve lying within the two Talek territories were
grazed year-round by livestock, and included broad vegetation types similar to
that in the Reserve, with somewhat reduced woody vegetative cover. Our 3rd
study group, the Mara River clan (28—43 members), defended a territory (31.0
kmz) >6 km from the Reserve border (Figure 1.1). We expected the more
isolated Mara River territory to be characterized by lower levels of tourist and
grazing pressure than either Talek territory, because the majority of tourist lodges

and all Maasai villages are situated outside the Reserve border (Figure 1.1).

Ecological comparison between clans

To assess the relative levels of human disturbance experienced by our
study clans, we conducted monthly comprehensive livestock censuses in the
Mara River territory (n = 18 months; Sept. 2002—Feb. 2004) and the Reserve
portions of both the Talek West (n = 18 months; Sept. 2002—Feb. 2004) and East
(n = 11 months; Sept. 2002—July 2003) territories. Censuses involved driving
throughout each territory to obtain complete head counts for sheep, goat and
cattle herds. Individual censuses were conducted at 2-h intervals throughout the
day, with an initial census conducted as livestock herds entered the Reserve in
the morning and a final census conducted as herds left the Reserve in late
afternoon. However, censuses at these time intervals were not necessarily

conducted on the same day; we completed one census during each 2-h interval

18

once a month, with an average of five censuses per month (all at different times)
in each territory.

As part of a larger ongoing study, behavioral observations of hyenas from
each clan were conducted on average 22 days per month throughout the study
period. Hyena observation “sessions” were conducted during all hours of the day
and night but occurred primarily during the early evening (1730—2000 h) and late
morning (0600—0830 h). To describe the timing and relative intensity of tourist
use of our clan territories we recorded all instances of tour vehicles approaching
hyenas during these sessions. These data were used to identify primary tourist
use periods during the 24-h period and to calculate the frequency of vehicle
approaches per hour of hyena observation during these periods in both the Talek
and the Mara River territories.

In addition to human disturbance we assessed other ecological variables
within each clan territory that might influence hyena activity patterns. We
characterized the availability of natural prey to hyenas in each clan by counting
all prey occurring along 1-km road transects in all three territories. Transects
were located in open grassland to facilitate counting and were placed evenly
throughout the territory, typically separated by less than 1 km (Mara River n = 24,
Talek West n = 11, Talek East n = 13; Figure 1.1). We counted all wild ungulates
within 100 m of each transect 2—4 times per month to estimate prey density
values for both territories. Because lions are an important source of mortality for

spotted hyenas (Kruuk 1972), we also recorded all observations of lions, either

19

alone or interacting with hyenas, and calculated average lion group size and the

relative frequency at which lions were observed within each territory.

Communal den use

The times of day at which females attend the communal den and nurse
cubs may be influenced by human activity. We used two different methods to
investigate whether the daily timing of den use in the two Talek clans (considered
together here due to low sample sizes within each clan) differed from that
observed in the Mara River clan. First, we utilized den observations lasting
several hours to monitor fine-scale timing of den-use. Observers arrived at the
communal den in the afternoon, before cubs emerged and before any other
hyenas were present, and recorded both the time of arrival at the den of the first
adult female or sub-adult, and the time at which den-dwelling cubs appeared
above ground. We compared the Mara River and Talek clans with respect to the
median time of first arrival and first cub appearance using a two-way test of
independence to test for equality of medians (Sokal & Rohlf 1995).

Second, we utilized information collected during shorter-term behavioral
observations at communal dens. Observers visited all active communal dens at
least once every two days during regular morning and evening observation
periods (see above). At each den visit, we conducted an initial scan to record
which hyenas, if any, were present at the den at that time. We conducted
subsequent scans of hyenas present every 10—15 minutes until a final scan was

conducted before we left the den. We used logistic regression, with time as a

20

single continuous predictor variable, to compare the Mara River and Talek clans
with respect to the influence of time on the probability of observing any hyenas at
the communal den. We performed separate regressions for each clan within
each observation period (morning and evening) and individual model significance

was assessed using a likelihood-ratio X2 test.

Hyena activity monitoring

We anesthetized 19 adult hyenas (11 females, 8 males) from two of the
three study clans (Mara River n = 10, Talek West n = 9) with
tiletamine/zolazepam (Telazol; W.A. Butler Company, Brighton, Michigan; 6.5
mg/kg) administered in a dart using a Cog-powered rifle (Telinject lnc., Saugus,
California), and fitted them with VHF radiocollars (Telonics lnc., Mesa, Arizona).
Radiocollared hyenas were from a wide range of social ranks. To describe
hyena activity patterns we utilized focal animal sampling with continuous
recording of behavior. These sampling events (hereafter “‘follows”) lasted 2—15 h
and were conducted at all times of day and night with the aid of night-vision
binoculars and infrared spotlights. Although Talek East hyenas were included in
analyses of timing of den use, they were not followed due to difficult terrain within
their territory. Because we were interested in comparing movement rates of
males and females without attachment to den sites, and because movements of
female hyenas are influenced by the need to return to den-dwelling young

(Boydston et al. 2003a), we only followed females without den-dwelling cubs.

21

Although we were unable to follow hyenas for complete 24-h periods due
to logistical difficulties, we documented the 24-h activity pattern of each individual
hyena by observing it during shorter follow segments that together generated a
composite 24-h cycle. We attempted to complete this cycle as quickly as
possible after its onset, with the average time necessary for completion being 31
days. Because 21% of the follow segments did not contribute to a composite 24-
h cycle (eg. due to hyena death or collar failure before cycle completion),
analyses utilizing data from all recorded follow segments (i.e. from composite
and incomplete 24-h cycles) are characterized by unequal sample sizes per
hour-long time block. All analyses requiring equal sampling throughout the 24-h
period utilize only data from composite 24-h cycles.

During each follow, we categorized the behavior of the focal hyena in
every minute as traveling, nursing, resting, socializing, feeding, engaging in
hunting, miscellaneous activity (standing, sniffing objects or the ground) or
interacting with other carnivore species. We considered all behaviors other than
resting and nursing as “active”. When individuals were out of sight, we
categorized animals that were clearly not resting (based on signal fluctuations,
auditory clues, or obvious location changes) simply as “active”, othenrvise we
recorded behavior and activity as unknown. We calculated the % time each
hyena was active and engaged in each behavior during each hour block of each
follow segment. Hour blocks with >20 min during which the animal was out of
sight, or activity could not be assessed, were not included in hourly behavior and

activity analyses, respectively. Follows were terminated when activity of the focal

22

animal could not be conclusively determined for >20 consecutive minutes or
when the focal animal was out of sight for >30 minutes within an hour block.
Activity during composite 24-h follows is presented as the % of the entire 24-h
period during which the animal was active. We also used composite 24-h follows
to describe hyena time budgets, with the % time spent engaged in each behavior
calculated out of the total minutes the animal was in sight during the 24-h period.
To determine whether hyena activity was affected by human activity, we
additionally compared individuals from the Mara River and Talek West clans with
respect to % of total activity exhibited during daylight hours (0700-1800 h), as
park regulations prevented public access to the Reserve between sunset and
sunnse.

We calculated the length of all bouts of activity and inactivity; we defined
the former as periods of at least 5 min of activity bounded on either end by 5
consecutive minutes of inactivity. We restricted analyzed bouts to those
beginning and ending between 1800—0900 h to exclude long periods of daytime
resting, and bout lengths were only calculated for follow segments lasting longer
than 4 h. Bouts lengths were minimum estimates because some follows ended
before bout completion. We identified cessation of daily activity as the time of
morning after which no more than 10 consecutive active minutes were observed,
and we identified onset of evening activity as the first afternoon time after which
at least 10 consecutive active minutes occurred.

Because social interactions and group formation may be facilitated among

distant hyenas through the use of long-distance vocalizations called whoops

23

(East & Hofer 1991b; Kruuk 1972), we also recorded all whooping behaviorby
focal hyenas to investigate its temporal distribution. As another measure of clan
social activity we recorded the number of hyenas with the focal hyena every 10
min during follows, and further noted group size changes as possible within each
10 min interval. We then calculated the mean minimum hourly number of
changes in group size using all composite 24-h follows.

To determine hourly movement rates, we recorded the locations of
followed hyenas every 10 min and calculated straight-line distances between
consecutive locations with Animal Movement Analyst (Hooge & Eichenlaub 2000)
and ArcView GIS 3.2 (Environmental Systems Research Institute, Redlands,
California). Very few locations required the use of telemetry as followed hyenas
were in sight, on average, for 98% of follow minutes. We then calculated
average hourly movement rates for individuals based on composite 24-h follow
cycles. Although we included only hour blocks with at least 4 locations per hour,
we collected, on average, 140 of 144 possible locations per 24-h follow cycle.

In addition to the use of composite 24-h cycles to calculate mean % time
active, and mean hourly movement rate for both daylight and the entire 24-h
period, we used all individual follow segments to perform additional activity and
movement comparisons, with data aggregated into 5 distinct diel periods. We
excluded data collected from 1100—1400 h because negligible amounts of activity
and movement were observed for any individual hyena during this time period.
We divided the hours of darkness into 3 periods of equal length (night1: 1900-

2300 h, night2: 2300—0300 h, night3: 0300—0700 h) and categorized the

24

remaining hours as morning (0700-1100 h) and evening (1400—1900 h), based
on average time of sunrise and sunset. We averaged hourly activity and
movement rates within each of these 5 periods for each individual and made
comparisons within time periods based on sex and clan. For comparisons of
movement rates between males and females based on these 5 periods, we
excluded all hour blocks during which the focal hyena engaged in a nursing bout
to control for female movement limitations due to nursing demands. However,
female movement rates based on 24-h follow cycles included follows during
which nursing behavior was observed.

We categorized social rank for adult females as high or low relative to the
median adult female rank. In analyses involving onset and cessation of activity,
the sampling unit was the follow segment but in all other analyses, the sampling
unit was the individual hyena. We averaged movement and activity data within
sampling periods (e.g. hour, morning period) for individuals that were followed
more than once during that period. Due to low sample sizes we compared mean
proportions between groups using a nonparametric Mann-Whitney U-test, and
compared frequency data using chi-squared tests.

Results in units of time are presented as medians because minutes are
not on a strictly continuous scale; all other results are presented as means t SE.
Listed p—values are two-tailed unless research hypotheses generated clear
directional predictions for group comparisons. For example, because we
predicted males would show higher rates of activity and movement, and spend

more time traveling than females, one-tailed tests were used for these

25

comparisons. Statistical comparisons for all analyses were considered
significant at a = 0.05. All statistical analyses were conducted using the software

package STATISTICA (StatSoft 2002).

RESULTS
We followed 19 different hyenas (11 females, 8 males) during 100 follow
segments for a total of 628 h. The average length of follow segments was 6.3 h
:l: 3.3 h (SD). These segments resulted in 22 composite 24-h follow cycles for 16
different hyenas (11 females, 5 males). Four females and two males were
followed for two 24-h cycles. Twenty-one follow segments did not contribute to a

composite 24-h cycle.

General activity

The daily pattern of hyena activity was largely nocturnal with no clear
peaks in activity throughout the night (Figure 1.2). In 16 follows of 11 different
hyenas the time of cessation of daily activity could be determined, and yielded a
median time of 0733 h (range: 0542—0918 h). The median time of onset of
activity in 27 follows of 17 different hyenas was 1834 h (range: 1443—2037 h).
The majority of activity observed during daylight hours occurred during the first
two hours after sunrise (Figure 1.2). Based on composite 24-h follows, in which
all hours were sampled equally, hyenas spent 31.5 i 2.7% of their time active.

During hours of darkness (1900—0600 h) hyenas spent 53.0 i- 4.1% of their time

26

 

 

 

60’ : s. . ‘ ”200 I
i V - . .-' 5
50 . «1000 E
2 v
:8- 1 i 2
l- . .l a
4; 40 800 n:
“e’ - e
i: 30I #600 g
a? . g
g 20- #400 g
E ' c
0:
10» , ~200 o
o - 2
0 .. . ". I.- . , . . I L 4 Lg 0

8 ° ° ° ° 8 8 8 8 ° 8 8

N 8 8 8 8 m o N <r 8 co 0

v- ‘- 1— ‘— N N O O O O O ‘—

Time (h)

Figure 1.2. Activity pattern and movement rates of radiocollared adult spotted ’
hyenas in the Masai Mara National Reserve, Kenya. Data are based on direct
observations during long-term (2—1 5 h) follows of males and females. Number of
individuals sampled varies per hour (n = 15—19). Plotted values for each hour
(eg. 1200 h) represent activity recorded in the following hour block (i.e. 1200—
1259 h). The black bar extends over hour blocks characterized by darkness

throughout the year.

27

active, and 96.2 i 0.9% of all activity occurred from 1800—0900 h. Only 9.7 i
1.2% of their active minutes occurred during daylight hours (0700—1800 h).

Movement rates showed the same general pattern as activity (Figure 1.2);
onset and cessation of movement were closely associated with sunrise and
sunset and the majority of daytime movement occurred in the early morning.
Using only composite 24-h follows, the mean movement rate for the entire 24-h
period was 584 i 64 m/h (range: 147—1185 m). Minimum nightly distance
traveled based on follows _>_8.5 h in length ()7 = 11.4 h) averaged 12.4 i 1.9 km
(range: 3.5-21.7 km).

Duration of active bouts from 1800—0900 h averaged 62 :t 6.2 min, with
48% of active bouts being less than 30 min in duration (Figure 1.3). There was
much individual variation in the timing and length of bouts of activity, and in the
total amount of time spent active (Figure 1.3). Bouts of inactivity from 1800—
0900 h averaged 53 i 4.5 min in length. The longest recorded active bout during
this period was for a male and lasted 383 min; the longest inactive bout was for a
female and lasted 257 min.

Group size change showed a bimodal pattern (Figure 1.4). Not
surprisingly, a similar pattern emerged from analysis of % time spent socializing,
indicating that the majority of clan interactions occur either between 1900 and
2300 h orjust after sunrise, between 0600 and 0700 h (Figure 1.4). Whoop
vocalizations were recorded from 1829 to 0706 h and were equally likely to occur

during all 2-h blocks from 1800—0800 h (f6 = 8.43, p = 0.209).

28

Figure 1.3. Bouts of activity (solid black bars) during 14 composite 24-h follow
cycles for 13 individual radiocollared spotted hyenas. Active bouts were >5
consecutive active minutes bounded on both ends by 5 consecutive minutes of
Inactivity. Pictured 24-h cycles were those for which the night was described by
no more than 2 follow segments. The letter “a” indicates a pair of follow cycles
conducted on the same individual. Start and end points of each follow segment
are indicated by a “”.v For example, the topmost follow cycle is composed of
three nonconsecutive follow segments conducted from 1730-0100 h, 0100-1000
h, and 1000-1730 h. Asterisks indicate hunts and fresh kills on which the focal

animal fed. Hours of darkness throughout the year fall within the shaded area.

29

 

 

  

 

 

 

 

 

I

 

 

 

 

 

 

 

C+

 

Talek West Clan

Mara River Clan

Figure 1.3.

30

 

 

 

3.0 I 9

+ o " 8 '
(I) 2.5 - 5". I
8) I: o 1'. "I. I . - 7 Em
C 5' ‘2, ‘1, II "'1 .E
2 2.0 “ ° 5 i. '2' 6 .N
N =' ‘1 ‘: . . '- ' : 'I 0
'5 15 .=' '-. '— .i - 2. w
0. I I‘ i: . a." 4 a)
3 I ‘ . E
h 1.0 ~ -" 3
(D 39
:H: 5' ° . 2 C
= r s
(U 0.5- s -
g ." .. I i I l ‘I 2

o . . I I X R‘-

0 0 4J4 L I 1 J; 1 L 1 1 A 4* 0

O O O O O C) O O O C o O

O O O C) O O C C O O O O

N V (.0 03 N N V' (.0 00 O

‘— 1— ‘- x— N N O O o O O 1—

Time of Day (h)

Figure 1.4. Mean number of group size changes observed and mean % time
spent socializing during follows of radiocollared spotted hyenas. Data are from
direct observations of 16 different individuals (11 female, 5 male) during 22
composite follows, each covering an entire 24-h period in nonconsecutive

segments. Other notation is as in Figure 1.2.

31

Sex differences

Comparison of male and female time budgets, based on composite 24-h
follows, revealed several differences. On average, males (n = 5) spent more
time traveling (U = 11.0, one-tailed p = 0.031, Figure 1.5). Males also spent
more time hunting than did females (U = 6.0, p = 0.015, Figure 1.5), but only 15
hunts were observed. Although focal females whooped during only 7 of 15
composite follows, focal males whooped during each of 7 composite follows.
Mean whoop rates for the 11 females and 5 males were 0.04 and 0.34 times/h,
respectively (U = 0.00, p = 0.002). Thus, males whooped 8.5 times more often
than did females.

Although both sexes started activity at the same time of day, males tended
to spend a larger portion of the 24-h period active (37.9 :I: 5.7%) than did females
(28.6 :t 2.7%, U = 16.00, one-tailed p = 0.097; Figure 1.6A, Table 1.1). The
greatest sex differences in activity occurred in the morning time block (26.7 i-
10.1% and 11.0 i 2.8%, U = 13.00, one—tailed p = 0.071; Table 1.1). Males
showed clear peaks in activity from 2200-2300 h, and from 0600—0700 h, during
which activity levels averaged close to 80% (Figure 1.6A), but comparable peaks
in female activity were not apparent.

Males tended to exhibit higher movement rates than did females over the
24-h period (U = 14.00, one-tailed p = 0.063; Table 1.1) with the most obvious
differences observed during the first half of the night (Figure 1.6B). Male
movement rates peaked at 0000—0100 h, but female movements did not show a

clear peak (Figure 1.6B). The maximum distance moved in a single hour was

32

80 r
- Males (n = 5)

 

 

‘8 Females (n = 11)
':
a)
o_
.r:
V
N
I..—
o
o\°
c
(0
a.)
E

*
U) U) > O) U) O) In
E E E 5 g E g
In {D 1:" = c

a: 2 <2 E e E

I— o' 8 I“

.9 <0 0

E ‘5'

Behavior

Figure 1.5. Mean % time spent engaged in various behaviors for 5 male and 11
female spotted hyenas during follows covering the entire 24-h period in
nonconsecutive segments. Miscellaneous activity included standing, and sniffing
objects or the ground. Traveling included both directed movement and wandering
behavior. Significant differences (Mann—Whitney U-test, p < 0.05) are indicated

by an asterisk.

33

90'

 

A —<>- Females
~--- Males
80 ~ 9
70» 53
o ' ‘-
..>. '
6 60’
<
0 .
g 50
'— 40
at?
r:
(U 30 .
m
E
20 -
10 »

 

 

 

O O O O O O O O O O O O

O O O O O O O O O O O O

N V (O 00 O N O N V to CO 0

\— Y— 1— v— N N O O O O o T‘

2200 r B —<>— Females
--'~ Males

 

2000i
1800»
1600-
1400-
1200»
1000»
800»
500r
400»

Mean Movement Rate (m/hr)

200 r

 

 

 

C) O O O O O O O O O O O

O O O O O O O O O O O O

N V (O (O O N O N Vt (O (O O

t- v- v- 1— N N O O O O O ‘—
Time (h)

Figure 1.6. A) Sex differences in mean % time spent active by radiocollared
adult spotted hyenas. Data are based on direct observation during long-term (2-
15 h) follows of males and females. Number of individuals sampled varies per
hour (females n = 9—1 1; males n = 5—8). Figure notation is as in Figure 1.2B.
Mean hourly movement rates of radiocollared adult spotted hyenas excluding
data collected during hour blocks when focal females nursed their cubs. Number

‘ of individuals sampled varies per hour (females n = 7—1 1; males n = 5—7).

34

0298 30:8 cém mgmanoo E0: Emu :0 2:0 nommm 0

£09.: >052>>Iccm2 8.5-9.0 Co 0:38.. :o ummmm a

 

 

000.0 3.0030 53.2 30.0 $50.00 £0030 .300 0 0w

000.0 2.0 :00 ca :03. {0.0 8.000. : :0 ES 0550.2

000.0 @0300: 885$ $0.0 exam? 6300.00 0292

000.0 @0050 300003 E0 850.3 £50.00 0392

«00.0 600030? 60033 30 8.3.00 3000.00 2092

02.0 6300 800000 02.0 330.0 8000.0, 05005
L a 2 L n. .2

20.5 9mm EmEm>oE amps.

w>zo< mEE. 0\0 amps.

 

.0 000 T000 05522 E 00500000 ”0292 5

000010000 ”9092 5 00000000 ”2092 5 0007003 055m. 030:9 00 9.0 000000 95 .008. 028$ .000002

Ems. 5me or: E mmcoE nouoam CV ofiES new 2): 20E Lou. 09m; EoEo>oE 050 33:00.. Ammv com—2 .: £an

35

4680 m by a male hyena and 4513 m by a female. All movement rate averages
are likely underestimates because extremely fast movement sometimes resulted

in termination of follows.

Effects of social rank

Because we found clan differences in some measures of activity (see
below), we restricted our analysis of social rank to those measures of activity that
showed no differences between the clans. We pooled females from both clans
resulting in use of 6 high-ranking (4 Talek, 2 Mara River) and 5 low-ranking (2
Talek, 3 Mara River) females. Based on composite 24—h follows, high-ranking
females were no less active (26.7 i 3.6%) than low-ranking females (30.9 i
4.2%) over the 24-h period (U = 11.00, one-tailed p = 0.233). In addition, high—
and low- ranking females showed no differences in the % time spent resting,
hunting or traveling (U > 7.0, p > 0.14), and they showed similar movement rates
during the 24-h period (U = 13.00, one-tailed p = 0.358). However, females of
lower rank spent more time feeding (2.2 i 0.4%) than did those of high rank (0.8

2': 0.3%; U = 3.0, p = 0.028).

Clan differences: Ecology

Monthly prey counts yielded estimates of prey density that did not differ
significantly among the clans (Mara River = 196.6/km2, Talek West = 210.8/km2,
Talek East = 181.9/km2; Kruskal-Wallis test: H248 = 2.17, p = 0.124). The

average size of lion groups seen within the territories of the two Talek clans (3? =

36

3.9 i 0.26) was no different than that of groups seen within the Mara River
territory (7 = 3.4 i 0.30, U = 6284.5, p = 0.108). In addition, 1-2% of observation
sessions included lions, either alone or with hyenas, in both Mara River and
Talek territories.

Illegal grazing of cattle, sheep and goats inside the Reserve occurred
nearly every day within the territories of both Talek East and West clans.
Monthly livestock censuses recorded an average of 991 cattle and 1038 sheep
and goats utilizing the area within the Talek West territory, and 515 cattle and
257 sheep and goats within the Talek East territory. Livestock grazing in the
Talek area, whether inside or outside the Reserve, typically occurred from 0900—
1800 h. In contrast, no livestock were ever seen in the Mara River clan’s
territory. Tour vehicles were present in all three territories primarily during only
two periods each day: 0630—0900 h and 1630—1900 h. During these periods,
researchers observed tour vehicles 4.8 times more frequently while observing

hyena groups in Talek (East and West pooled) than in the Mara River territory.

Clan differences: activity and movement

Because males and females were not sampled equally between clans,
and because we demonstrated differences in activity and movement rates
between the sexes, comparisons of activity and movement between the Talek
West and Mara River clans include only data from females. Female hyenas from
the Talek West clan were followed 30 times for 206 total hours (2 = 6.9

h/segment) and Mara River females were followed 29 times for 176 hours (3? =

37

6.1 h/segment). Using composite 24-h follows, Mara River females (n = 5) were
no more active (27.3 i 4.0%) than were Talek West females (29.7 i 3.9%, n = 6)
over the 24-h period (U = 0.715, p = 0.715; Table 1.2). However, differences
between clans were apparent in the timing of activity, with Mara River females
showing more activity than Talek West females in the evening (one-tailed p =
0.025; Figure 1.7A) and during daylight hours (one-tailed p = 0.034; Figure 1.7A).
In addition, the evening onset of socializing was delayed for Talek females
relative to Mara River females (Figure 1.78). Although females from both clans
showed an early evening peak in % time spent socializing, that peak occurred 3
h earlier in Mara River than in Talek (Figure 1.78). Comparisons of 24-h time
budgets showed no differences between the two clans in % time spent engaged
in any specific behaviors (U > 7.0, p > 0.17). Additionally, no differences were
observed in overall movement rates or timing of movements during the 24-h

period between females from the two clans (Table 1.2).

Clan differences: communal den use

We conducted 17 evening long-term observations at communal dens in
Mara River and 28 in Talek. Den-dwelling cubs were first seen at communal
dens earlier in the evening and at a wider range of times in Mara River (median =
1716 h; range = 1540-1854 h) than in Talek (median = 1844 h, range = 1738-
1906 h; p < 0.001). In addition, time of arrival at the communal den of either an

adult female or large subadult was earlier and more variable in Mara River

38

0065 30:8 :ém 0:009:00 E0: 0000 :0 2:0 0000m o

00:_0>.Q 00:00-0:0 0

0000.3 >0:~_:>>I::0_>_ 00 0:300: :0 0000m 0

 

 

 

0000 800000 800000 0:0 80:00 8.0000 000 c 40
100.0 60000 £0009 £0000 3.00.0 a. 00.00 059.00
000.0 @3000 500000 .0000 8,000.0 8,000.0 0550.2

000.0 :55? $000000 0.30 $050.00 $030 0292

0000 630002 A00 :00: 000.0 3.000. B 8. F 00.00 0592

000. 0 80000000 80000000 000. F 3.00000 A00 00.00 0092
90000 80000 A000: n000.0 :. :00 :00 2.: 05020

.0 E 0.2 a 2: ms.
EEV 000m E0E0>0§ :00_2 0>=0< 06:. 0\0 :005.
.0

oowvioono 0_ £9.00 0:0 .. 0_:0._. :_ 00 000000 0E2. .0>:0x 02000”. 05:02 0.0.2 000.). 05 55:5 0:0.0A12v00zm

0:02 :0 9?: 000>> x0_0._. 05 .050 E0: 00:03 000000 0.00:00 :00 0000: E0E0>0E 0:0 33:00 ER :005. .NA 0.00...

39

 

 

 

 

 

 

 

 

8° ' ----- Mara River Clan
A —o— Talek West Clan
70-
060*
.2
“'0'
<50-
g
p40»
a?
:30:
0
0
220»
10[ -, '
.I
°séssésésée§§
93933988883889
IO' B
U
16 i -+- Mara River Clan
E14, —0— Talek West Clan
.E
.912-
0
0
(010.
3
i: 8'
a? 6,
:
a |
g 4’
2’
~"I
_-- - A
0 C O O O O O O o o 8 O O
O O o O O o O O o O C
ezssa~ssasse

Figure 1.7. The mean % time spent A) active and B) socializing by radiocollared
adult female spotted hyenas from two different social groups (clans). Data are
based on direct observation during long-term (2—15 h) follows. Number of
individuals sampled varies per hour (Talek West n = 5—6; Mara River n = 4—5 for

activity and n = 3—5 for socializing). Figure notation is as in Figure 1.2.

40

(median = 1654 h, range = 1323—1853 h) than in Talek (median = 1834 h, range
= 1730—1908 h; p < 0.005).

We recorded 3300 scans (morning: 1353, evening: 1947) at active
communal dens In Talek East and West clans and 1562 scans (morning: 759,
evening: 803) in the Mara River clan. Logistic regression indicated that time had
a significant influence on the probability of Talek hyenas being present at the
communal den in both the morning (Log-likelihood 121 = 11.778, p < 0.001) and
evening (Log-likelihood X21 = 53.022, p < 0.001) periods; scans were more likely
to reveal no hyenas late in the morning and early in the evening (Figure 1.8). In
contrast, time did not influence hyena presence at communal dens during either
the morning (Log-likelihood X21 = 1.130, p = 0.288) or evening (Log-likelihood x21
= 0.704, p = 0.401) observation periods in the Mara River clan (Figure 1.8).
Therefore den use, as documented by both observation methods, appeared to
begin later in the evening and end earlier in the morning in the Talek clans than
in the Mara River clan, further reinforcing the apparent evening phase delay

suggested by observed differences in the timing of activity and socializing.

DISCUSSION
General activity
Overall activity levels of Mara hyenas corresponded closely with those
reported for Crocuta in South Africa (Kalahari Gemsbok NP: 31.5% of 24 h,

53.0% of nighttime—Mills 1990; Kruger NP: 27.5% of 24 h—

41

.5
01

A - Mara River Clan
- Talek East and West Clans

-‘ N (a) U #
01 O 8 O 01 O

% of Scans with No Hyenas
3

 

1730-1759 1800-1829 1830-1859 1900-1929 1930-1959

4° l - Mara River Clan

B - Talek East and West Clans

% of Scans with No Hyenas

 

 

0600-0629 0630-0659 07000729 0730-0759 0800-0829
Time Block

Figure 1.8. Percent of A) evening and B) morning scans at active communal
dens in which no hyenas were present during each of 5 half-hour blocks for the

Mara River Clan and the Talek East and West Clans.

42

Henschel 1986). Mara hyenas also spent amounts of time resting (70.7%) and
traveling (19.9%) that were similar to those reported for Kalahari Crocuta (69%
inactive, 23.6% foraging). However, Kalahari hyenas traveled more than twice
as far each night (7 = 27.1 km/night) than did Mara hyenas. This difference may
reflect the much larger clan territories observed in the Kalahari (7 = 1095 kmz),
where prey densities are far lower than in Kenya (Mills 1990). The nightly
distance traveled by Mara Crocuta (7 = 12.4 km) matched more closely that of a
followed female in Ngorongoro NP (7 = 10.1 km—Kruuk 1972), where clan
territories are more similar in size to those of Mara clans (Ngorongoro NP: Y =
23.8 ka—Honer et al. 2005).

As in previous studies, we observed activity onset to occur around sunset,
with the majority of daytime activity occurring early in the morning (Kruuk 1972;
Mills 1990). Like Henschel (1986), but in contrast to Kruuk (1972), we found little
evidence of consistent peaks in activity through the night, likely resulting from a
large degree of individual and nightly variation in activity. However, peaks in
socializing and group size changes offer insight into the fission-fusion sociality of
Crocuta. The majority of socializing within the clan, particularly by females, is
conducted soon after onset of activity each evening. Groups formed during this
period often remain relatively stable through the night until just after sunrise,
when groups undergo additional reshuffling before the daytime rest period. We
found the frequency of whooping activity to be constant throughout the night, as

was also shown for males at dens in the Serengeti (East & Hofer 1991a).

43

Although activity graphs depicting an average activity pattern (as in Figure
1.2) are useful for describing general patterns, they fail to depict episodes of rest
that may occur frequently and unpredictably throughout the night. Our study
reports the first records of duration of active and inactive bouts for this species
and indicates a highly punctuated pattern of nightly activity, with frequent bouts of
inactivity interspersed among active bouts of highly variable duration. Individual
differences in the timing and length of active bouts were remarkable, as has been
observed in other large carnivores (Garshelis & Pelton 1980; Theuerkauf et al.
2003b). However, occasional and unpredictable group behaviors such as border
patrols, lion-hyena interactions and inter-clan conflicts influenced timing and
levels of nightly activity within individual follow segments, and replication within
individuals would be necessary to adequately assess individual variation in
activity. Previous studies of some felids have noted a detectable reduction in
activity for multiple days after kill events (Puma concolor—Beier et al. 1995, L.
rufus—Schmidt 1999). However, whereas felids may take several hours or days
to consume a carcass, hyenas feed with remarkable speed (eg. 18 kg
food/h/hyena—Kruuk 1972) and commonly feed in groups, thus reducing the

likelihood that single kill events would significantly influence patterns of activity.

Sex and rank differences
Because follows were logistically difficult, many of our analyses suffer
from small sample sizes and few replicates within Individuals, so factors such as

weather and prey abundance may have influenced individual activity and

44

movement estimates. We attempted to follow individuals exclusively during clear
weather, and across a wide range of prey abundance conditions, but we were
otherwise unable to control for these factors. Although previous studies have
shown seasonal variation in hyena movements due to prey fluctuations (Hofer &
East 1993a; Trinkel et al. 2004), these studies were conducted in areas where
prey abundance in some regions decreased annually to near-zero values
whereas the Reserve supports relatively stable resident ungulate populations
throughout the year. Furthermore, when pooling females from both study clans,
we found no correlation between local prey densities and hyena movement rates
or activity, and two females that were followed in both periods of high and low
prey density showed very similar activity rates (unpublished data). However, our
sample sizes prevented us from specifically controlling for these factors and our
results should therefore be interpreted with due caution.

We predicted, based on sexual selection theory, that the significant
positive relationship between male reproductive success and time spent with
receptive females (East et al. 2003; Szykman et al. 2001) would demand higher
rates of movement and activity in males than in females. Indeed, like Henschel
(1986), we found male spotted hyenas tended to be more active and to travel
further than females during a 24-h period. Here, differences were most
pronounced in the hours just after sunrise and sunset.

The African lion displays fission-fusion sociality similar to that of spotted
hyenas, with group composition being both temporary and unpredictable

(Schaller 1972). Although male lions of unknown or nomadic social status have

45

been shown to range more widely than resident females during the night
(Hemson 2003; Stander 1997), comparisons between males and females of the
same resident pride have not been performed. The fission-fusion society of
Crocuta is also remarkably similar in many respects to that of some primates,
including spider monkeys (Ate/es spp.—Chapman 1990; Symington 1990) and
chimpanzees (Pan troglodytes—Lehmann & Boesch 2004). Although
chimpanzees generally show no sex differences in their time budgets, males
travel faster and cover greater daily distances than do females (Doran 1997). In
spider monkeys, the fact that males range more widely, travel faster, and spend
more time traveling than females within the shared group territory has been
suggested to be the result of males monitoring females and patrolling territory
boundaries (Shimooka 2005). Because participation in territory defense and
border patrolling behavior is no more common in male than female Crocuta
(Boydston et al. 2001 ), we believe the sex differences we observed reflect male
monitoring of females rather than territory boundaries.

Although low-ranking female spotted hyenas range more widely than high-
ranking females (Boydston et al. 2003a; Honer et al. 2005), we found no rank-
related variation in movement rate or activity levels among the females followed
here. We also found time budgets to be remarkably similar between high and
low-ranking females with the exception of more time spent feeding by low-
ranking females. Although high-ranking hyenas spend more time feeding than
low-ranking clan members in competitive feeding situations (Frank 1986b), we

suspect that low priority of access to kills forces low-ranking individuals to rely on

46

lower quality food (e.g. bone) and may force them to engage in longer, more
frequent feeding bouts over time. One would similarly predict low-ranking
females to hunt more often than high-ranking females, and although our sample
size for hunts was too small to detect rank effects, previous research has

demonstrated this trend (Holekamp et al. 1997b).

Clan differences

Female hyenas in the Talek West Clan showed less daytime and early
evening activity than did females in the Mara River Clan. In addition, the range
of times utilized by Talek East and West females when attending the communal
den was small relative to that of Mara River females, with Talek den use starting
later in the evening and ending earlier in the morning. Various ecological
differences might potentially have caused the observed differences in use of time
between study clans, yet the Talek and Mara River territories differed very little
except with respect to the intense daily exposure of Talek hyenas to the
presence of tour vehicles and livestock.

Our data suggest that tour vehicle activity alone could not account for
observed differences between clans with respect to hyena activity and den use.
Despite an often heavy volume of tour vehicles from 0630-0830 h, when hyenas
are generally active but before cattle enter the Reserve, we found no reduction in
activity or movements of Talek females relative to Mara River females during this
period. Additionally, despite constant tourist presence in Talek from 0630—0900

h, den use by Talek hyenas progressively decreased as the start of the grazing

47

period approached. During the evening onset of hyena activity, when both tourist
use and livestock activity was high in Talek, hyena activity was reduced, and
both den use and peaks in socializing were delayed relative to those observed
among Mara River females. Tour vehicles, while abundant in both Talek clan
territories, do not pose a direct threat to hyenas and have been present in this
ecosystem for more than three decades. Maasai herdsmen, however, represent
a direct threat to hyenas, as humans are an important source of mortality for this
population, second only to lions (Watts & Holekamp in review). Additionally,
hyenas appear to perceive herdsmen as a threat because hyenas often flee from
guarded cattle herds, whereas cattle left unattended by herders are not avoided
(unpublished data). This response has also been noted among African wild
dogs (Fuller & Kat 1990). Taken together, these observations suggest that cattle
grazing, but not tourism, affected temporal distribution of activity among Talek
hyenas.

Although the influence of human disturbance in the form of livestock
grazing on the activity patterns of other carnivore species has not been
investigated, similar activity shifts to those observed here have been described in
response to a wide variety of anthropogenic disturbances with predictable
temporal schedules. Carnivores that typically show some degree of diurnal
activity have been found to increase or completely rely upon nocturnal activity
when faced with the threat of hunting mortality (Andelt 1985; Kitchen et al. 2000)
or harassment resulting from human recreational activities (Beckmann & Berger

2003a; Olson et al. 1998), except where spatio-temporal avoidance is possible

48

(Theuerkauf et al. 2003a). Similarly, daytime movements and activity have been
reduced or eliminated among carnivores living in urban, or human-dominated
landscapes (Ciucci et al. 1997; Lucherini et al. 1995; McClennen et al. 2001;
Riley et al. 2003).

Although direct fitness costs might be expected to result from reduced
diurnality in species adapted for daytime or crepuscular activity (e.g. cheetah,
bears, coyotes), such direct costs are not expected in hyenas. Their night vision
is believed to be as good as their day vision (Kruuk 1972), and variation in
hunting success with time of day has not been reported. However there may be
indirect fitness costs in this species when daytime activity is eliminated. For
example, diurnal activity may reduce risk of predation on hyena cubs by lions,
which are active almost exclusively at night (Schaller 1972; Stander 1992).
Similarly, little daytime activity by predators was suggested to explain diurnal
activity exhibited by female wolves with young cubs in an othen~ise nocturnal
group (Vila et al. 1995). In addition, a reduction in the range of times available to
females to nurse cubs at dens could potentially reduce overall numbers of
nursing bouts, increase energetic stress on both cubs and mothers, and promote
social conflict over den access. Although we did not observe enough nursing
bouts or lion-hyena interactions in the three clans to address these hypotheses,
long-term demographic data indicate no recent reductions in cub survival or clan
size in the Talek area (Watts 8 Holekamp in review), suggesting that these
indirect costs, while potentially important, have not reduced fitness in these

hyenas.

49

We have shown that human disturbance in the form of livestock grazing,
can alter patterns of activity and den use in this species. While it remains
unclear what fitness costs might ultimately be incurred by spotted hyenas from
disturbance-based alteration of normal activity patterns, documentation of
behavioral changes such as these among large carnivores is critical to the
development of our understanding of human-carnivore interactions and the
extent to which carnivores can adapt to human presence. Changes in activity
patterns or movements can potentially be used as early behavioral indicators of
the extent and severity of human disturbance and may have other unforeseen
consequences. For example, reduction of daytime activity by large carnivores
might negatively influence monitoring efforts based on sighting data (Caro et al.

1998), or reduce the frequency with which tourists can observe these animals.

50

CHAPTER TWO

ECOLOGICAL AND ANTHROPOGENIC INFLUENCES ON SPACE USE IN THE
SPOTTED HYENA (CROCUTA CROCUTA)

INTRODUCTION

It has become increasingly clear that continued presence of large
carnivores is often critical to proper ecosystem function (Berger 1999; Crooks 8
Soulé 1999; Ripple et al. 2001; Terborgh 2001). However, large predators are in
global decline due to a combination of habitat degradation and direct human
persecution (Weber 8 Rabinowitz 1996; Woodroffe 2001), and mortality
associated with “edge effects” threatens the persistence of even protected
carnivore populations (Woodroffe 8 Ginsberg 1998). Given the increasing
fragmentation of carnivore habitat in an expanding matrix of agriculture and
urbanization, there is a critical need to understand the ecological factors,
including various forms of human disturbance, that influence space use decisions
by large carnivores. These data may be critical, for example, in guiding reserve
design and management (Caro 8 Durant 1995), and in assessing extinction risk
and resilience in response to disturbance or change (Arcese et al. 1997).

Because resources available to carnivores in areas influenced by human
activity inevitably vary in quality, responses to human disturbance represent
tradeoffs between optimal resource use and decreased risk of persecution (Gill 8
Sutherland 2000). With this perspective, the overall importance to carnivores of
resources within disturbed environments should influence their response to
disturbance, as well as the costs associated with this response. It is therefore

essential to determine not only the factors influencing space use decisions, but

51

also the costs and trade-offs associated with these decisions. For example,
alterations in carnivore space use resulting from human disturbance may reduce
exposure to humans and therefore the threat of human-caused mortality, yet
access to critical prey or habitat resources may be compromised. Quantification
of these costs and trade-offs will provide an increased ability to forecast the
consequences of disturbance of varying degrees, and identify the costs to
carnivores of functional habitat losses associated with human activity.

The majority of Africa’s large carnivore species have experienced recent
and often dramatic reductions in range, due largely to habitat conversion
(Ginsberg 8 Macdonald 1990; Mills 8 Hofer 1998; Nowell 8 Jackson 1996).
Although the spotted hyena (Crocuta crocuta) is no exception, it is currently listed
as Lower Risk (IUCN 2006), and this species occurs in relatively stable
populations throughout much of its historic range, a fact attributed to its
behavioral and ecological plasticity. For example, spotted hyenas are found in
habitats ranging from desert to rainforest, and occur in social groups (clans)
ranging in size from 10 (Mills 1987) to 80 (Kruuk 1972). As is the case with many
African carnivores, hyenas frequently share the landscape with human
pastoralists; therefore livestock are a common landscape feature for many hyena
populations. Whereas in many systems worldwide livestock left unguarded in
pastures are unlikely to serve as a disturbance to carnivores, livestock
throughout much of East Africa are closely guarded, and herders pose a direct

threat to carnivores. Hyenas have been shown to alter their patterns of activity in

52

response to livestock grazing (Kolowski 8 Holekamp 2007) and to reduce use of
intensely grazed areas over time (Boydston et al. 2003b).

Multiple studies have demonstrated carnivore avoidance of areas
characterized by intense human activity (wolves—Thurber et al. 1994; mountain
lions—Van Dyke et al. 1986; coyotes—Gese et al. 1989; bears—Mattson 1990;
Olson 1994; Reinhart 1990), yet few have attempted to quantify the trade-offs
involved in disturbance response decisions (but see Gibeau et al. 2002). In
addition, few researchers have investigated the influence of livestock on
carnivore space use patterns (Chavez 2006), particularly where the presence of
herds is associated with potential for direct persecution from herdsmen.

The primary objective of this study was to identify the factors influencing
patterns of hyena space utilization. By describing these relationships
concurrently for hyenas in both disturbed and undisturbed environments within
the same ecosystem, we attempted to isolate the influence of livestock grazing
on space utilization. In addition, by monitoring a number of potentially important
ecological variables concurrently, we attempted to identify the potential costs to
hyenas associated with this disturbance, in terms of their ability to optimize use
of territory resources. Finally, we investigated the mechanisms by which hyenas
exposed to livestock grazing may adapt to this disturbance.

Boydston et al. (2003b) showed that large portions of the territory of a clan
of hyenas (“Talek” clan), in the Masai Mara National Reserve, Kenya (hereafter
the Reserve) that were heavily used by hyenas from 1988—1990, were avoided

from 1996-1998. Ungulate and livestock sampling demonstrated that avoided

53

areas were characterized by both the highest prey densities and the most
intensive livestock grazing in the territory. Consequently, the presence of
livestock within the territory may have serious fitness implications for Talek
hyenas in that effective habitat loss due to grazing might be associated with loss
of access to critical prey resources. In addition Boydston et al. (2003b)
demonstrated increased use of densely vegetated habitat corresponding in time
with dramatic increases in livestock grazing within the territory.

Before human presence can be causally linked with recent observed
changes in Talek hyena behavior, baseline knowledge of resource use patterns
in the absence of livestock grazing must be described to allow separation of
direct and indirect human influence from influences of the natural environment
(Arcese 8 Sinclair 1997; Caro 1999a). The Mara River (MR) clan, which defends
a territory free from livestock grazing in the center of the Reserve, served here as
a baseline control. The results of Boydston et al. (2003b) suggest that Talek
hyenas, over time have increased their use of vegetative cover, potentially as a
refuge from grazing herds and pastoralists. Their results further suggest that
Talek hyenas, due to the presence of livestock, were unable to utilize prey
resources within their territory in an optimal fashion. We therefore hypothesized
that the presence of livestock grazing as a frequent and intense disturbance
results in adjusted space use decisions by hyenas that sacrifice optimal resource
use, for reduced threat of persecution. Specifically, we tested the following
predictions: 1) densely vegetated habitats should be less heavily used by MR

hyenas than by Talek hyenas, 2) prey distribution should be a more effective

54

predictor of space use by MR hyenas than Talek hyenas, and 3) probability of
use of a given area by Talek hyenas should be negatively associated with
intensity of use by livestock.

Herds grazing in the Reserve typically enter the park between 0800-1000
and leave by nightfall (Boydston et al. 2003b). Because livestock grazing is
predictable in time and space, hyenas may restrict use of grazed areas to times
of day when grazing is absent, and therefore minimize effective loss of habitat
within the territory. This may be particularly true if avoided areas are resource-
rich (Boydston et al. 2003b), because habitat available to hyenas is constrained
by territorial boundaries and agonistic interactions with hyenas in neighboring
clans. Such spatio-temporal shifts have been seen in a number of carnivores,
with individuals increasing use of disturbed areas during daily periods when the
disturbance was minimal or absent (Ciucci et al. 1997; Gibeau et al. 2002;
Machutchon et al. 1998; McLellan 8 Shackleton 1988). We therefore tested the
additional prediction that hyenas exposed to livestock should modify their
temporal patterns of space use to minimize effective habitat loss. Specifically,
we expected that: 1) Talek hyenas would be found in areas of high livestock
disturbance more frequently during periods in each 24-h cycle when grazing is
rare or absent than when grazing is frequent and, 2) hyenas would be found
further from areas of the most intense livestock grazing during grazing periods,

than during disturbance-free periods.

55

METHODS

Study area

This study was conducted from September 2002—April 2004 in the Masai
Mara National Reserve in southwestern Kenya. The Reserve supports a large
diversity of resident ungulates including gazelles (Gaze/la thomsonii and G.
grantr), impala (Aepyceros melampus), topi (Damaliscus lunatus), and giraffe
(Giraffe camelopardalis). In addition, from August to October small resident
populations of wildebeest (Connochaetes taun'nus) and zebra (Equus burchellr)
are joined by large migrant herds from Tanzania. The Reserve consists primarily
of rolling grassland habitat and scattered shrubland with riparian forest along the

major watercourses.

Study populations

We monitored individual hyenas from two clans in the Reserve. The Mara
River (MR) clan defended a territory near the center of the Reserve (Figure 2.1)
and included 28—38 individuals (7 = 8 adult females, 7 = 5 immigrant males).
The Talek West (TKW) clan defended a territory along the northwestern border of
the Reserve (Figure 2.1) and contained 47-55 hyenas (3? = 11 adult females, 3? =
7 immigrant males). Due to its location along the Reserve border, and its
proximity to a number of pastoral villages (Kolowski 8 Holekamp 2006), the TKW
clan territory is subjected to daily livestock grazing pressure. By contrast, no
livestock were ever seen grazing within the territory of the MR clan during our

study period.

56

Hyena locations

We anesthetized 29 adult hyenas (18 females, 11 males) from our two
study clans with tiletamine/zolazepam (Telazol; W.A. Butler Company, Brighton,
Michigan; 6.5 mg/kg) administered in a dart using a COz-powered rifle (Telinject
lnc., Saugus, California), and fitted them with VHF radiocollars (Telonics lnc.,
Mesa, Arizona). The 12 hyenas (7 females, 5 males) collared in the MR clan had
functional collars for an average of 396 d (range = 98—578 d) during our study
period. The 17 collared hyenas (11 females, 6 males) in the TKW clan were
monitored for an average of 391 d (range = 66—608 d). Radiocollared hyenas
spanned a wide range of social ranks in both study clans. We made attempts to
locate each collared individual on a daily basis. Most radiotracking was
conducted during the morning (0530—0900 h) and evening (1730—2000 h), but
additional locations were collected throughout the day and night. The majority
.(66%) of all tracking locations were based on direct sighting of hyenas (MR: 72%,
TKW: 60%). When visual confirmation was not possible, we were able to localize
the radio signal to an area typically < 200m2; therefore spatial resolution of
tracking locations was high in all cases. To avoid temporal autocorrelation in
tracking locations, all locations for the same individual were separated by at least
1 hour, as that is a conservative minimum estimate for the time it takes an
individual hyena to cross its entire territory (White 8 Garrott 1990). However,

individuals were rarely tracked more than twice a day.

57

 

  

    
 

 

  

  

 

 

   
 

i
Masai Mara A
National Reserve , ”as“ ”8‘“
1 0 1 2 Kilometers _ National Reserve
{If
. er g,
(5.0!. ..
s "
. -. , “x. J
Mara River \ ,.- .9,
T I s I. '0‘ I I: -, 9k. RIP
erntory :' \ .- e,
I rLQC'HWKTI a]. ‘ ‘1' — ‘1 03
{7 fit- ./ r “ ’:.: E 1‘ ,- 0 0
€ \ .."* j, \ \ 5 “'0 OO %
‘: m" /'~ 0° 0 o
2 ,. - I y - :' ‘ ab
‘ T \ j , .0; 5 T I“ \
‘ fl -- f 4' f 5 : l
l I; ,4: 5 g : I
-, \r‘s ‘/ ~. "I. - . " 1: \ Talek WBSt .5 ,1
“Fax!!- "0’ .4 .3"); if. I I ‘s:\l Territory 2: I: \‘\\\—-~——..‘ {-k
I I”: I i" : .2 \ ‘ \“ “1" ’
”7"" \ I, ’ I t I : \I. \' \I‘ ‘ "v:
\ T \ T I \ x 3 ‘ ‘ I
. i . ’l I \ x
t i ‘-. ', / T‘ \
l IUIIIIIIII I I ‘-.
1‘ : A) ”‘0’ ' ‘3‘ \
I
"'.\
0.. \

Figure 2.1. The location of the territory boundaries of the two study clans.

Ungulate transects are shown with barred lines and Maasai villages within 2 km

of the Reserve are shown with open circles. Note the extension of the Talek

West territory outside the Reserve boundary.

58

Hyena clan territories

We identified territory boundaries of the MR clan based on a combination
of territorial behavior and radiolocations of adult females. An initial Minimum
Convex Polygon (MCP) was created using all recorded locations of communal
dens (see “Dens” section below) and territorial behavior, which included
boundary patrols, aggressive interactions with neighboring clans, and latrine sites
(see descriptions of these behaviors in Henschel & Skinner 1991; Kruuk 1972;
Mills & Gorman 1987). This polygon was then manually adjusted to align with
natural boundaries known to limit hyena movement (i.e. the Mara River). Finally
the polygon was expanded to match a 95% fixed-kernel home-range contour
(Powell 2000; Worton 1989) created using all tracking locations of radiocollared
adult females from June 2001-April 2004, where this contour extended beyond
the MCP. This process was repeated for the TKW clan, but female locations
here were limited to those collected after 1 April 2002, when a long-term fission
event in this clan was conservatively estimated to have been complete. This
fission resulted in creation of the Talek West and Talek East clans from a single
“Talek” clan. The territory of the MR clan was 31 .Okm2 and was based on 2407
tracking locations; the territory of the TKW clan was 28.4km2 and was based on

1967 tracking locations.
Livestock distribution

To document the spatial distribution of livestock grazing, we regularly

conducted comprehensive livestock censuses in the Reserve portions of the

59

TKW territory. Censuses involved driving throughout the territory to obtain
complete counts of sheep, goat and cattle herds. Herds containing less than 500
head were estimated to the nearest 50 animals; herds containing >500 were
estimated to the nearest 100 animals. Individual censuses within months were
conducted at 2-h intervals throughout the day, with an initial census conducted
as livestock herds entered the Reserve in the morning and a final census
conducted as herds left the Reserve in late afternoon. However, censuses at
these time intervals were not necessarily conducted on the same day; we
completed one census during each 2-h interval each month, with an average of
five censuses per month (all at different times). Livestock are corralled at night in
the villages outside of the Reserve. Based on observed grazing patterns we
identified the hours between 0900 to 1800 h as “grazing hours” (GH) and the
hours from 1900 to 0800 h as livestock-free hours (LFH). We considered the
hours from 1800 to 1900 h and 0800 to 0900 h as transition periods between
grazing and livestock-free hours, because the timing of herd movement into and
out of the park was somewhat variable.

We recorded a point location for the center of all herds using car-mounted
GPS units by approaching the herds at close range, and estimated the shape
and size of the area in which the herd was grazing. Herd shapes were then
drawn in a GIS and filled with randomly located points according to the recorded
herd size, with each cow represented by a single point location. Because
livestock data were based on complete censuses rather than sampling transects,

we did not use interpolation methods to create livestock distribution surfaces.

60

We instead created a utilization distribution grid surface for the entire territory
using fixed-kernel methods with values of each grid cell weighted according to
the actual density of livestock locations observed within and around each cell
(Seaman & Powell 1990; Worton 1989). Grid surfaces were created for each
month of the study as well as for larger time intervals, depending on the
analyses, by pooling livestock locations (from censuses at all 2-hr intervals)
across the period of interest. The smoothing parameter for each grid surface
kernel function was initially determined using the reference bandwidth value
(Worton 1995). If the resulting grid surface indicated areas of zero utilization
between herds deep in the Reserve and herds entering the Reserve, the
smoothing parameter was increased iteratively by 20 until at least low livestock
utilization was indicated between late afternoon and early morning herd
locations. This was done to ensure that grid surfaces reflected the fact that
herds deep in the Reserve must have traversed the space between their grazing
location and the Reserve border when not observed. Grid cell lengths for
created surfaces varied with the number and density of livestock points, but
ranged from 40—100m. Final grid cell values, referred to hereafter as livestock
use values (LUVs) ranged from 0-100, with higher numbers indicating higher
grazing intensity. Although all livestock in the area were counted, analyses

utilized only herds recorded within the territory boundaries of our study clan.

61

Ecological Variables
Dens

Two types of dens, natal and communal, are utilized by spotted hyenas
(East et al. 1989; Kruuk 1972; Mills 1990). Hyena cubs are typically born in a
natal den at which no other litters reside. After a period of 1—5 weeks (East et al.
1989; Kruuk 1972), cubs are moved to the clan’s communal den, which is often
used concurrently by several litters of varying ages. Cubs in the Reserve reside
at the communal den, on average, for the next 7-8 months (Holekamp et al.
1996), at which time they begin to follow their mothers throughout the territory
and gradually gain independence. Typically a single communal den is used
within a territory at any given time (Kruuk 1972; Mills 1990), but den locations
may change frequently (e.g. every 1.5 months average for Talek Clan—Boydston
et al. 2006).

When females have cubs at natal dens, they typically spend large portions
of their time there. Once cubs are moved to the communal den, mothers spend
less time at the den but often visit it at least twice a day to nurse their cubs
(Kruuk 1972). When females have den-dwelling cubs they are found, on
average, closer to the communal den than they are at other phases of their
reproductive cycle (Boydston et al. 2003a), and their core-areas tend to be
centered on communal den locations (Boydston 2001). In addition to functioning
as a refuge for vulnerable cubs (East at al. 1989; Holekamp & Smale 1998), the
communal den also serves as a center of social interaction and sub-group

assembly for the clan (Holekamp et al. 2000; Kruuk 1972; Mills 1990). Thus, in

62

addition to females with cubs, it was expected here that all other clan members
would, to some extent, exhibit space use patterns influenced by the location of
the communal den. We recorded the coordinates of all communal dens utilized
by MR and TKW hyenas during the study period, as well as their dates of usage.
Water

Water was not expected to be a limiting resource here, as permanent
water sources are found throughout the Reserve, and hyenas living elsewhere
within this ecosystem show little dependence on free water (Kruuk 1972).
However, the placement of water features may indirectly influence hyena
movements, for example, by providing cool daytime resting sites or by influencing
preferred vegetative characteristics. Therefore, we obtained maps in vector
format of all permanent and seasonal streams digitized from aerial photographs
(Boydston 2001) and, when necessary, streams were digitized directly in the field
with a hand-held GPS unit.
Prey Distribution

Spotted hyena ranging patterns are influenced by the local distribution of
prey resources (Hofer & East 1993a; Mills 1990; Trinkel et al. 2004), Within the
Reserve, hyenas have been found to use areas of higher prey density more
frequently than areas of lower prey density, except when disturbed by intensive
grazing (Boydston et al. 2003b). Here, we characterized the availability of natural
prey to hyenas in each study clan by counting all prey occurring along multiple 1-
km road transects in both territories. Transects were located in open grassland

to facilitate counting, were placed evenly throughout each territory, and were

63

typically separated by less than 1 km (Mara River n = 24, Talek West n = 14;
Figure 2.1). Because the primary prey species of Crocuta are ungulate grazers,
we assumed that shrubland and riparian forest habitat, in which no prey transects
were located, supported the lowest prey densities in each territory. We counted
all wild ungulates within 100 m of each transect 2—4 times per month. Each
kilometer of transect was divided into thirds, with the number of prey counted
along each third associated with a midpoint for each section. Therefore, each km
of transect was assigned three points, each with its own corresponding prey
count for each sampling event. This allowed estimation of variability and spatial
correlation in prey densities over short distances, as well as over the longer
distances between transects. Each point was assigned the mean number of prey
counted on the corresponding transect section (0.333 km * 0.2 km = 0.067km2)
over the period of interest (“prey value”). For each month, as well as for longer
periods of analysis, a prey distribution grid surface was then interpolated from
these data points using inverse distance weighting (power function = 2). Final
grid surfaces were composed of relatively large 500 m x 500 m cells covering
each territory to represent the course resolution of the field data. Each cell
therefore represented the estimated prey value at that location based on transect
points within and around each cells according the inverse distance weighting

function.

64

Vegetation

Vegetation may influence hyena space use in a variety of ways.
Vegetation classes may be effective indicators of other important ecological
features such as diurnal resting spots or prey distribution. Additionally, dense
vegetation may offer important refugia for hyenas in disturbed habitats, as
indicated by Boydston et al. (2003b). Therefore, we created a vegetation map for
each study clan territory based on a combination of three data sources:
vegetation type reference points collected in the field, a 15m-resolution
panchromatic Landsat 7 ETM+ image, and a 30m-resolution 6-band Landsat 7
ETM+ image. Both Landsat images were collected on 4 Feb 2003 and were geo-
referenced using previously digitized road intersections and streams. Vegetation
reference points were recorded in both territories (Mara River n = 899; Talek
West n = 1257) during the last 11 months of the study period, during which all
tracking locations were associated with a particular vegetation class at the time
of collection. A vegetation map for each territory was digitized by hand in a GIS
with three brad vegetation class: riparian forest, shrubland, and open grass plain
(Figure 2.2). Shrubland was defined as dense woody cover, the vast majority of
which was Croton or Euclea bushes approximately 2m in height. Riparian forest
often included this shrub layer but was distinguished by the presence of a tree
canopy. We assumed that these 3 classes were sufficient to describe the cover
types available to hyenas, given the low diversity of vegetation types in this area
and the overwhelming dominance of grass plains in both territories (~ 80% of

territory area for each clan, see Figure 2.2).

65

 

Mara River Territory

 

Talek West Territory

 

 

 

 

Figure 2.2. Vegetation class maps for the territories of the Mara River and Talek
West hyena clans. Permanent and seasonal streams are shown in blue, and the
three vegetation classes are grass plain (yellow), shrubland (brown) and riparian

forest (green).

66

Logistic regression models

We used logistic regression models to identify relationships between
observed hyena space use patterns and the suite of ecological variables
described above. Logistic regression modeling is a common statistical method
used to calculate resource selection probability functions, whereby the probability
that a location or area is used by an animal is a function of a set of habitat
variables associated with that location (Manly et al. 2002). The estimation of
these functions when using radio-telemetry data is often based on the
comparison of a sample of locations used by an animal (i.e. tracking locations)
with a sample of locations available for use (Manly et al. 2002). Characteristics
associated with locations used by animals are then compared to those at
available locations with “used” vs. “available” modeled as the binary dependent
variable. Variables identified in significant models are those that effectively
differentiate between “used” and “available” locations, and therefore represent
the factors influencing resource selection by a particular species. We used this
methodology to identify variables associated with resource selection in spotted
hyenas by comparing ecological features associated tracking locations of spotted
hyenas (“used”) to those associated with 5000 random locations identified within
each clan territory (“available”).

Each hyena tracking location was associated in a GIS with specific values
of the ecological variables described above. In order to model the influence of
the den location, given that it was not constant throughout the study period, all

modeling was restricted to periods during which den location did not change, or

67

was moved less than 1 km. In the latter case, the average UTM coordinates for
all dens occupied within that modeling period were used to represent one den
location for the period. All locations, both “used” and “available”, were then
assigned a distance to the average den location for each modeling period. In
addition, each location was associated with a prey value and, for TKW only, a
livestock use value, based on the grid cell value underlying the location. These
cell values were based on separate livestock and prey distribution maps created
for each modeling period. We then assigned each “used” and “available” location
to a vegetation class based on the digitized vegetation maps (Figure 2.2). Finally,
for each location, we calculated the distance to the nearest stream of any size,
the nearest permanent stream and the nearest vegetative cover. Cover was
defined here as all areas not classified as grass plain.

All spatial analyses were performed using either ArcView GIS 3.2 or
ArcGlS 9.2 (Environmental Systems Research Institute, Redlands, California)
with the help of a number of program extensions. Animal Movement Analyst
(Hooge & Eichenlaub 2000) was used to create all MCPs and fixed kernel home
ranges, and to calculate distances between point locations and specific habitat
features. Association of point locations with the grid cell values of prey and
livestock distribution grid surfaces was performed using the Grid Tools extension
(Jenness 2006) and Spatial Analyst. Inverse-distance weighted surfaces for prey
distributions were created using Spatial Analyst in ArcView and ArcGlS. Satellite
images were processed and georectified in Erdas Imagine 8.6 (Leica

Geosystems LLC, Norcross, Georgia).

68

Model creation procedure

Separate logistic regression models were created for each modeling
period with the following predictor variables: vegetation class, distance to den,
distance to nearest stream, distance to nearest permanent stream, prey value,
and LUV (for Talek West models only). Because vegetation class was a
categorical variable, it was modeled using grass plain as a reference category.
Each model compared variables associated with random locations (“available”)
with the same variables measured for hyena tracking locations (“used”), using
only hyena locations collected during that modeling period. Because reproductive
state was found to influence ranging patterns of adult females hyenas (Boydston
et al. 2003a), model comparisons between clans may be biased depending on
the relative numbers of tracking locations for females with and without den-
dependent cubs. Therefore all female locations were categorized according to
whether or not each female currently had cubs residing at dens. We then
excluded locations for females with den-cubs that were recorded at either natal
or communal den locations. Additionally, since random locations had to be
identified within an area deemed “available" to hyenas (here the territory), we
similarly limited hyena locations for these analyses to those falling within territory
boundaries. This resulted in a loss of less than 4% of tracking locations in each
clan. In the TKW clan, since livestock data were only available for locations
inside the Reserve, we also limited locations for analysis in this clan to those
occurring inside the Reserve, resulting in a further loss of 3.6% of all tracking

locations for TKW hyenas.

69

For each modeling period, all possible logistic regression models using all
recorded variables were compared using information-theoretic methods
(Burnham & Anderson 2002). These methods utilize Akaike’s Information
Criterion (AIC) values to incorporate both model parsimony and model fit to
assist in selection of appropriate descriptive models. All models within 2 NC
points of the optimal model were investigated and considered to have empirical
support based on the data (Burnham & Anderson 2002). We identified the
predictor variables included in all models in this initial subset, and ran a final
logistic regression model that including only these variables. The significance of
variables in each final model was based on the Wald statistic, and overall
significance of each model was tested using likelihood ratio tests.

We calculated odds ratios for each variable in each model to compare the
relative influence of ecological variables among modeling periods for the same
clan, and between the clans themselves. In logistic regression, an odds ratio
represents the change in the odds of the dependent variable being “true” (in this
case being identified as a “used” rather than an “available” location), associated
with a one-unit change in a predictor variable (e.g. 1m for distance variables).
Here, we report modified odds ratios (Long 1997) that represent the % change in
probability associated with a biological meaningful change in each predictor. For
distance measures we used an increase of 100 m, for prey values an increase of
5 units (estimated density), and for livestock use values an increase of 10 units

(% probability of livestock use). We used the average modified odds ratio for

70

individual variables over all models within clans to qualitatively assess the overall
influence of independent variables in the models on space use in each clan.

To investigate the potential influence of time of day on the relative
importance of ecological variables, we used the above methods to run separate
logistic regression models within each modeling period using hyena locations
collected either during grazing hours (GH: 0900—1800 h) or livestock-free hours
(LFH: 1900—0800 h). For this analysis and those described in the next section,
we assumed that prey distribution did not change significantly between the GH
and LFH hours. During radiotracking sessions we noticed no obvious changes in
prey distribution between day and night periods and nighttime assessments of

prey distribution were logistically impossible.

Analysis of spatio-temporal shifts in utilization

We used two different methods to investigate the possibility that hyenas in
the TKW clan utilized spatio-temporal avoidance of livestock to prevent effective
habitat loss. We first inquired whether livestock use values associated with hyena
locations collected during GH were lower than those collected during LF H, as this
would indicate increased use of disturbed areas when the disturbance is absent.
We also asked whether the proximity of hyena locations to areas characterized
by the most intensive livestock use differed between GH and LFH.
Tracking Locations and Livestock Use Intensity

Because livestock use intensity varied from month to month, we first

investigated the relationship between the distributions of livestock and hyenas on

71

a monthly basis. We created livestock utilization distributions, as described
above, for each month of the study period and associated each hyena tracking
location, based on date of collection, with the corresponding LUV for the location.
We included only tracking locations falling within both Reserve and clan territory
boundaries, and excluded seven months when <20 tracking locations were
collected during either GH or LFH. First, we compared monthly LUVs of hyena
locations to the average LUV of all cells within the Reserve portion of the TKW
territory. We considered hyenas to have avoided livestock when the 95%
confidence interval for the mean hyena location LUV did not overlap with the
overall territory LUV. Within each month, we then calculated the average LUV
of hyena locations collected during GH and LFH. lf spatio-temporal avoidance
was occurring, we expected LUV values to be higher for locations collected
during LF H. We compared mean values in a paired t-test using months as
sampling units. The latter analysis was then repeated using individual hyenas as
sampling units instead of months. For these comparisons, we assigned LUVs to
hyena locations based on a livestock utility distribution created using livestock
data from all months during which the individual was monitored (using all herds
plotted during this period). We included only individual hyenas that were tracked
>20 times during both GH and LF H. Included hyenas were all monitored
continuously for at least 18 months. We used independent sample t-tests to
compare average LUVs for locations collected during GH and LFH for each
hyena, and a paired t-test, using hyenas as sampling units, to assess the overall

difference between LUVs of GH and LFH locations.

72

Tracking Locations Relative to Livestock Core Area

The second method used to determine whether TKW hyenas avoided
livestock examined locations of tracked hyenas relative to the area(s) in which
the most intense livestock use occurred throughout the study period (livestock
core area: “LCA”). The LCA was identified as the area within the 50% contour of
a fixed-kernel livestock utility distribution based on all herds recorded in the TKW
territory throughout the entire period during which each hyena was radiotracked.
Each hyena location was assigned a distance to the LCA, and was also
categorized as being within or outside this area. We recorded the proportion of
GH and LFH hyena tracking locations that fell inside the LCA, as well as the
average distance to the LCA for both sets of locations. Proportions and mean
distances were calculated on a monthly basis, excluding seven months where
<20 locations were collected during GH or LFH. Mean distances were compared
using t-tests and proportions were compared using ftests. Overall differences
between GH and LFH locations in the distance to the LCA were compared using

a single paired t-test and months as sampling units.

RESULTS
Logistic regression models
Modeling Periods
We identified three modeling periods for the MR clan, which averaged 161
d in length (range: 80—274 d), during which the location of the den did not change

more than 1km (Table 2.1). We identified 4 such periods for the TKW clan that

73

Table 2.1. Periods identified in the study period during which den location was
stable in the Mara River (MR) and Talek West (TKW) clans. Logistic regression
modeling of resource use was based on data collected only during these periods.
Levels of prey and livestock for each period are represented by the mean # of
ungulates/km of transect and the mean # of livestock animals counted per

census drive during each period, respectively.

 

Period Dates Days Hyenas (f,m) Tracks Prey Livestock
MR1 10/03/02 — 12/21/02 80 9 (6,3) 387 43.5 —
MR2 12/23/02 — 4/30/03 129 10 (6,4) 596 16.0 —
MR3 5/04/03 — 2/01/04 274 9 (5,4) 1249 54.3 —
TKW1 11/01/02 — 4/06/03 157 12 (6,6) 578 37.9 2266
TKW2 7/28/03 - 11/02/03 98 11 (6,5) 407 58.2 1046

TKW3 12/26/03 — 2/06/04 43 12 (8,4) 235 30.4 1078

TKW4 2/07/04 — 4/30/04 85 12 (7,5) 464 56.0 615

 

74

averaged 96 d (43-157 d; Table 2.1). After excluding a small subset of locations,
as described above, the mean number of tracking locations per period was 744
in the MR clan (range: 387-1249) and 421 locations in the TKW clan (range:
235—578; Table 2.1). The average proportion of MR locations collected during
LF H and GH for these modeling periods was 50.8% and 26.8%, respectively.
The average proportion of TKW locations collected during LFH and GH was
46.5% and 21.7%, respectively.

Univariate Comparisons ( “Used” vs. “Available” Locations)

Prey distribution grid surfaces indicated variability in both the distribution
and abundance of prey among modeling periods for both clans (Table 2.1, Figure
2.3 and 2.4). Livestock utility distribution grids showed similarly high variability in
intensity of grazing among modeling periods in TKW (Table 2.1, Figure 2.5), but
the spatial distribution of livestock, in contrast to that of prey, appeared stable
over time (Figure 2.5). The monthly mean number of livestock counted per
census in the TKW territory was 1386 :l: 181 (range: 106-3160).

We identified clear differences between used and available locations in both
clans based on simple univariate comparisons. In fact, all measured ecological
variables, both continuous (Figure 2.6, Table 2.2) and categorical (Figure 2.7,
Table 2.3), showed significant differences between used and available locations.
Hyenas in both clans were located significantly closer to the den than expected in
all modeling periods (Figure 2.8 and 2.9), and, with the exception of one
modeling period (TKW1), hyena locations were associated with higher prey

values than were available locations. Univariate comparisons showed no

75

 

 

 

 

Figure 2.3. Prey distribution grid surfaces (cell length = 500 m) for each of the 3

modeling periods for the Mara River clan. Darker colors indicate higher estimated
prey densities. The communal den location used during each period, and the
clan territory boundary, is indicated on each map. Note that prey distributions

indicated outside territory boundaries are not accurate.

76

 

 

Figure 2.4. Prey distribution grid surfaces (cell length = 500 m) for each of the 4

modeling periods for the Talek West clan. Darker colors indicate higher
estimated prey densities. The communal den location used during each period.
and the clan territory boundary is indicated on each map. Note that prey

distributions indicated outside territory boundaries are not accurate.

 

 

 

0 2 Kilometers
:5

 

 

 

 

Figure 2.5. Livestock utility distribution grid surfaces for each of the 4 modeling
periods in the Talek West clan. Darker colors indicate higher intensity of use by
livestock. The communal den location used during each period, and the clan

territory boundary is indicated on each map.

78

8O

'IMR

Mean % Deviation from Available

 

 

 

Den Stream Perm Stream Cover LUV Prey Value

Figure 2.6. The mean % difference between values of continuous variables
measured at hyena locations (“used") and those measured at 5000 random
locations (“available”) within the respective clan territories. A % difference was
calculated for each of 3 modeling periods in Mara River (MR), and 4 periods in
Talek West (TKW); the mean of these values is indicated above. The first 4
variables represent distances; therefore negative % deviations indicate hyena
selection for proximity to these features. Variable descriptions, the data from
each individual model, and the associated statistical comparisons are shown in

Table 2.2.

79

 

 

80

EN 6.:on tn €.:~..o omm 8:3: 5.“: 3:me 8m $38 38 28:88 :5:
313:? 8.... 34:3 omw 8:8 53 Eve: 8m Amer? 58 5:92 :5:
0.5 3:08 we. advemm cum 8:? BE €382 mom 63%: 58 3:32 «5:
92 3%.? 3 8.05.0 omm 3:9: 88 $939 Ba $33..” ©va 6988 :5:
---:..._....A..w..smm.-------H--------.. ........ em----m._.v.m.m.----mallmmvma ..... may.“...N.---.....WM...N.--.A.......HH.,..MW ...... my.
3 21:5 - - 8m a:8~ :8 3:98 9% 6:5 Bow $383 was.
02 3.05.8 - - 8m smvmmm 5m 8:28 9%. £53 £8 8:39 Es.
__m>< 8: __m>< 8: __m>< 8: __m>< 8: __m>< 8: __m>< 8:

o:_m> >96 >3._ Lo>oo Emotm E51 EmmEm con.

.28 5 2m _0 $3 :0 comma 5:6 ._.02

En “or: 329 2:0 .m>onm 89.68: mm @2388 >05 0cm. momsmcoo x0982. E0: 8620 moomtsm 26 :0 women
295 mm:_m> 35 use 93.: mo:_m> om: x0069: .5508 Lo>oo 639.com; 69mm: 6cm AEmobm Ema: Emobm EocmELoa
69mm: .AEmobwv Emobm 69mm: 9: .Eoo: con _m::EEoo 65 2 AmLBmE EV woos—820 3:29.86 05 “commie.
839.? 8.5 E: 9:. .220 96.: as; 6.3 Em Es: 62m em: 9: :0 85:9 as 55? $5265.: 25:82

E065“: ooom ncm .A..Umm:..v 20:80. 9.209: mac? L2 8282 $33.? msosczcoo Lo wo:_m> :0 £08: :85. dd 632.

220 r - MR
2 TKW

    

180 ~

140 >

100 ~

60'

20-

Mean % Deviation from Available

 

 

Shrubland Riparian Forest Grass Plain
Vegetation Class

Figure 2.7. The mean % difference between the proportion of hyena locations
and the proportion of 5000 random locations (“available”) within the respective
clan territories that were located in each of 3 vegetation class. The % difference
was calculated for each of 3 modeling periods in Mara River (MR), and 4 periods
in Talek West (TKW); the mean of these values is indicated above. The data
from each individual model and associated statistical comparisons are shown in

Table 2.3.

81

Table 2.3. The percent of hyena radiolocations (“used”) and 5000 random

locations (“available”) within the territories of the Mara River (MR) and Talek

West (TW) clans that were associated with each of three vegetation classes.

Periods indicated on the left represent time periods during which hyena locations

were collected (see Table 2.1 for period descriptions and location sample sizes).

 

 

 

 

Shrubland Riparian Forest Grass Plain
Used Avail 12, Used Avail f, Used Avail 12,
MR1 13.7 7.9 16.11* 7.0 13.2 1263* 79.3 78.9 0.04
MR2 16.1 7.9 45.42* 13.9 13.2 0.22 70.0 78.9 24.73*
MR3 12.0 7.9 2169* 13.9 13.2 0.41 74.1 78.9 13.62*
TKW1 38.9 16.8 163.39* 4.5 2.3 10.42* 56.6 80.9 180.51*
TKW2 72.2 16.8 695.93* 2.9 2.3 0.74 24.8 80.9 661.16*
TKW3 50.2 16.8 166.93* 1.7 2.3 0.34 48.1 80.9 147.63*
TKW4 38.8 16.8 134.57* 1.9 2.3 0.22 59.3 80.9 119.74*

 

' Significant [2 test statistic at o = 0.05.

82

 

MR1

 

 

 

 

 

 

 

 

 

 

Figure 2.8. Locations at which adult male and female Mara River hyenas were

radiotracked within their territory during 3 modeling periods (see Table 2.1).

Locations at the communal den for females with den—dwelling cubs were

excluded. Permanent and seasonal streams are indicated along with the

communal den (orange triangle) used during each period.

83

 

 

TKW4

 

 

 

 

 

Figure 2.9. Locations at which adult male and female Talek West hyenas were
radiotracked within their territory and within the Reserve during 4 modeling

periods (see Table 2.1). Locations at the communal den for females with den-
dwelling cubs were excluded. Permanent and seasonal streams are indicated

along with the communal den (orange triangle) used during each period.

84

avoidance of livestock use areas relative to available locations, with data from all
modeling periods indicating either hyena selection for areas of higher livestock
use, or no difference in livestock use values between TKW used and available
locations. MR hyena locations were consistently closer to streams in general,
and further from permanent streams than were available locations. However,
locations for TKW hyenas did not show a consistent relationship with water
features except for an apparent selection for areas close to permanent streams.
With respect to vegetation class, hyenas in both clans showed selection for
shrubland and corresponding avoidance of grass plains, relative to their
availability, though this appeared to be a much stronger tendency in TKW.
Multivariate Analysis - Mara River Clan

Final logistic regression models in each of 3 modeling periods for the MR
clan generated similar results (Table 2.4, A1). The most important predictor
variable in all 3 final models was distance to the den. Except for exclusion of
distance to vegetative cover from the MR1 model, final models included all
predictor variables. All 3 models were highly significant based on the likelihood
ratio test (f > 636.0, p < 0.0001).
Multivariate Analysis - Talek West Clan

The relative importance of predictor variables varied, sometimes
dramatically, among modeling periods for the TKW clan (Table 2.5, A2, A3, A4,
A5). Distance to the den was an important predictor variable in all 4 final models,
but was not consistently the most important. Vegetation class was selected as

the first or second most influential variable in all 4 models. Livestock use value

85

Table 2.4. Logistic regression results for Mara River modeling periods based on

hyena locations collected during all times of day and night.

 

Model Variable Estimate SE Wald Stat. P
MR1 Dist to Den - 0.0009 0.00006 243.10 <0.0001
Dist to Stream - 0.0032 0.00034 87.89 <0.0001
Prey Value +0.0508 0.00926 30.06 <0.0001
Dist to Perm +0.0005 0.00015 10.66 0.0011
Habitat 9.96 0.0069
Shrub/and +0.3688 0.13065 7.97 0.0048
Riparian F. - 0.4533 0.15484 8.57 0.0034
MR2 Dist to Den - 0.0007 0.00005 190.56 <0.0001
Dist to Cover - 0.0017 0.00027 38.03 <0.0001
Prey Value +0.0245 0.00513 22.71 <0.0001
Dist to Stream - 0.0015 0.00034 19.88 <0.0001
Habitat 15.00 0.0006
Shrub/and +0.2337 0.09964 5.50 0.0190
Riparian F. +0.1179 0.10161 1.35 0.2459
Dist to Perm +0.0004 0.00014 9.72 0.0018
MR3 Dist to Den - 0.0009 0.00004 596.97 <0.0001
Dist to Perm +0.0010 0.00009 139.23 <0.0001
Dist to Cover - 0.0016 0.00023 51.79 <0.0001
Dist to Stream +0.0017 0.00025 44.78 <0.0001
Prey Value +0.0259 0.00519 24.91 <0.0001
Habitat 19.97 0.0005
Shrubland - 0.0480 0.08199 0.34 0.5582
Riparian F. +0.3147 0.08145 14.93 0.0001

 

86

Table 2.5. Logistic regression results for Talek West modeling periods based on

hyena locations collected during all times of day and night.

 

Model Variable Estimate SE Wald Stat. P
TKW1 Dist to Den - 0.0014 0.00010 201.27 <0.0001
Habitat 123.15 <0.0001
Shrub/and +0.6059 0.10049 36.35 <0.0001
Riparian F. - 0.0439 0.17060 0.07 0.7968
Dist to Perm +0.0006 0.00012 22.94 <0.0001
Prey Value - 0.0430 0.00914 22.17 <0.0001
LUV - 0.0162 0.00533 9.22 _ 0.0024
imamnsgy‘vala """"" .1 56735 """" 6 352592 """" 1’ 5331' """ lofoooi
Habitat 152.46 <0.0001
Shrubland +0.8041 0.12789 39.53 <0.0001
Riparian F. +0.0489 0.22787 0.05 0.8300
Dist to Den - 0.0009 0.00012 57.76 <0.0001
Dist to Stream - 0.0023 0.00036 40.11 <0.0001
___________ 90393901---"_::9_-99_Q§____-"2999.12"______J_Q-21________-9-9_919.
TKW3 Dist to Den - 0.0007 0.00011 48.79 <0.0001
Habitat 34.75 <0.0001
Shrubland +0.7455 0.19350 14.84 0.0001
Riparian F. - 0.4801 0.35379 1.84 0.1748
Dist to Perm - 0.0005 0.00011 21.94 <0.0001
Dist to Cover - 0.0025 0.00059 18.58 <0.0001
___________ 9i§3.t_9§t_'§€m_____t9_-99_1_1________9-.999_~°2‘}____-____1Q-§2.______--9_-9911
TKW4 Habitat 116.15 <0.0001
Shrubland +1.1694 0.13708 72.78 <0.0001
Riparian F. - 0.9341 0.23941 15.22 <0.0001
Dist to Den - 0.0004 0.00004 88.47 <0.0001
Dist to Perm - 0.0008 0.00009 77.15 <0.0001
Dist to Cover +0.0018 0.00031 32.13 <0.0001

 

87

was selected as a predictor in only one final model (TKW1), with increasing
livestock use associated with lower probability of hyena use. Interestingly in this
model, increases in prey values reduced the probability of hyena use of territory
locations, and prey value was not even selected for 2 of the 4 final models. All 4
final models were highly significant based on the likelihood ratio test ((2 > 340.0,
p < 0.0001).
lnterclan Comparisons

A comparison of model results between the MR clan (where no livestock
grazing occurred) and the TKW clan, where livestock grazing was common,
revealed some important similarities and differences. Distance to the den was
consistently an important predictor variable of hyena use in both clans, and
showed a similar degree of influence on model prediction over all models (Table
2.6). However, although distance to the den was consistently the most important
predictor of MR hyena locations, other variables, particularly habitat type, were
more important in some TKW models. Because we could not exclude all
locations from females with den-dwelling cubs, there was the potential for these
females to contribute unequal amounts of locations to each model data set, and
lead to spurious conclusions regarding the influence of the den. However, the
average % of locations contributed by females with den cubs to TKW models
was 24.3 (range: 9.7-56.3%), and 19.3 (range: 10.3-27.1%) for MR models,
indicating that differences between clans in the influence of the den were not due
to clan differences in the reproductive states of monitored females. In addition, in

the TKW modeling period during which more than half the locations came from

88

 

 

v. m+ m- w- F+ a- nmN+ _.I.+ mm: o_.+ n+ ©+ v- w- w- cam—2
5+ mN+ 2+ 0 3. o oov+ mom+ NNN+ Nm+ or- o m+ o. v. v>>v_._.
mm- or- mm. 0 5+ 3+ om: o _.:.+ o o o a- m- n- m>>v:.
o o 0 mm. 9.. ON- er+ ~©+ VN: 5+ ©m+ 31 o a- m- N>>v_._.
o o o m- N_.+ o 2.: Nm+ mw+ mm- ON. 9.- or- 2... m7 _.>>x._.
w...--mfl--mc...--------mm-.w.c...--ms.. .......... a. m. ..... WM... ........ w.--a.. an. ....... H--m---m ............ a?
o or- m? ma: m- 9.. 5+ w- m- o 2+ 3+ 9.. w- m- mg).
3.- or- m..- 2.. m7 .3- m:+ o ©N+ 2+ 2+ m_.+ or- m- m- was.
0 o C 0v. 3- RN- nm+ o mv+ o om+ mN+ or- 3. w- 5:2
IO In... =< IO In... __< IO I”: __< IO In... =< IO In: =< Eve—2
hm>oo Emmbm nacm >91 .50

.5505 2m 33.2.9 882% __m 62 .o E m:_m> m umcmfimm 295 £258 .95

5.. 8823 6: mm_nm:m> .o:_m> $5 5 $855 :56 m 98 co>oo 923%? .o .Emobm 69mm: ion 9: 9 cosmic

E 38.05 E on: m. 53> 3508mm m_ 55 .chE m 3 :33: mm 8:3:me 9:8 8:80. a Io 3:5305 E 3:20 o\o $5

9865 385:2 .2 88:82 “15:52 02.78059: Em .E 25788 ”192:9. 9.85 Ea: 2268. 9:8:

cam AWN cam in 2an u..__<..v «62:80. 9.23: mam? __m :o comma 8:88 225.: .2 «.039 £60 8:605. .mN min...

89

females with den cubs, the model indicated den location was less important than
prey distribution and vegetation class.

Differences between clans in the relative influence of additional ecological
variables were also evident. Although MR hyenas consistently avoided
permanent streams (of which there were few) but selected sites closer to streams
in general, the relationship between these watercourses and TKW hyena
locations was unclear. Similarly, although prey was selected as a significant
positive predictor for MR models, the TKW models indicated an inconsistent
relationship between hyena space use and prey distribution (Table 2.5 and 2.6).
Although members of both clans showed selection for shrubland vegetation,
comparison of odds ratios indicates a much stronger influence of this vegetation
class on hyena space use decisions in the TKW clan (Table 2.6). The
identification of a location as shrubland increased the estimated probability of
identification as a “used” location by an average of 22% in MR, but by 135% in
TKW.

Temporal Differences in Model Results

Separate models based on locations collected during GH and LF H (Tables
A.6 - A.19) indicated some obvious temporal variation in space use patterns by
hyenas in both clans. For example, variables related to vegetative cover
generally had a stronger influence on locations collected during GH than during
LF H (Table 2.6). Because much of the vegetation in the Reserve is associated
with water features, distance to streams is almost certainly related to selection for

cover as well as moist/muddy resting areas. In the MR clan, distance to stream

90

was a stronger predictor variable for locations collected during GH (Table 2.6),
when hyenas spend most of their time resting in shaded muddy areas, but the
relationship between space use and water features was not as clear in the TKW
models. In both clans, the importance of shrubland was greatly increased during
GH, which undoubtedly represents, at least in part, selection for shade cover.
However, selection for shrubland vegetation remained during LFH in TKW, but
disappeared during LFH in MR. Although distance to cover was consistently
lower at hyena locations than at random locations in both clans (Table 2.2),
distance to cover was not a consistently important predictor of hyena space use
based on multivariate modeling in either clan and did not show clear temporal
trends in model influence (Table 2.6). This may be because selection for
vegetative cover was described more effectively by the vegetation class variable,
leaving little variation in space use to be described by the distance to cover
variable. In the MR clan, prey values generally had a stronger influence on
locations collected during LFH than during GH (Table 2.6), which likely reflects
increased hunting behaviour during nighttime hours. However, this trend was not
consistently evident for TKW locations. Although variable in its importance
among models within clans, the influence of the den appeared consistent overall
between GH and LF H (Table 2.6).
Spatial Correlations

Ecological variables were not completely independent, and investigation of
these intercorrelations is essential for a comprehensive understanding of

modeling results. In particular, correlations between den location and the relevant

91

ecological variables may lend insight into hyena selection of den locations within
the territory. Distance to the den location was negatively correlated with prey
values for 6 of the 7 modeling periods (Table 2.7), indicating that dens tended to
be located in areas with higher than average prey values. This trend can be seen
clearly when den locations are overlaid with prey distributions (Figures 2.3 and
2.4). In fact, the high correlation between den location and prey distribution in
model TKW4 (the only r-value between any variables that was >0.70) is likely the
reason that prey value was not selected as a predictor in this model. In the TKW
clan, the relationship between den location and livestock grazing intensity was
less clear, but it is remarkable that in 2 of the 4 models for this clan, correlations
indicated the location of den sites in areas subjected to higher than average
usage levels by livestock (Table 2.7, Figure 2.5). Finally, because the ultimate
goal of the modeling was to investigate costs associated with livestock as a
disturbance, we were interested in the relationship between prey values and
livestock use intensity. Here also correlations did not show a clear relationship
between the variables, with 2 of the 4 models indicating that areas of high
livestock use were associated with high prey values (Table 2.7). For all
correlations p-values are not presented because they would be heavily

influenced by the extremely large sample sizes (> 5000 points).

92

Table 2.7. Pearson correlation coefficients between 3 ecological variables used
in logistic regression modeling of space use by members of the Mara River (MR)
and Talek West (TKW) clans. Variables listed are the distance to the communal
den location (Dist to Den), the livestock use intensity value (LUV) and the prey
value. Correlations are based on all locations used for modeling during each

period (i.e. 5000 random locations plus all tracking locations for the period).

Period Dist to Den * Prey Value Dist to Den * LUV LUV * Prey Value

 

MR1 - 0.24 _ _
MR2 - 0.58 — _

MR3 +0.22 — —

'T'kw} """""""""" - 034035 """""""""""" i EH? """"
TKW2 -0.14 -040 +0 34

TKW3 - 0.05 +0 02 - 0 04

TKW4 -0.80 +013 -018

 

93

Spatio-temporal avoidance

We did not identify a clear relationship, within months, between livestock
use values (LUVs) associated with hyena tracking locations and the average
LUV of the TKW territory (Table 2.8). Data from only 3 of 13 months indicated
that hyenas used areas characterized by lower intensity of livestock use than that
available overall in their territory. Interestingly, during 4 of the 13 months, hyena
locations had higher LUVs than the average territory value, indicating selection
for areas used by livestock. This overall lack of avoidance of livestock use areas
is supported by the lack of significance of LUV as a predictor variable in 3 of the
4 logistic regression models discussed above.

Despite an overall lack of avoidance, there is still an opportunity for
hyenas to modify their use of disturbed areas on a temporal level. However, we
found little evidence of this in TKW hyenas. Livestock use values did not differ
between hyena locations collected during GH and locations collected during LFH
on a monthly level (n = 13 months; t= 1.11, p = 0.289 Table 2.8), or at the level
of the hyena (n = 11 hyenas; paired t= -1.22, p = 0.25; Table 2.9). At a courser
scale, hyenas tended to be found in the area of the most intense livestock
grazing (LCA) less often during GH than during LF H, but these differences were
not significant (Table 2.10). In addition there was no difference between LF H
and GH in the average distance to the LCA at which hyenas were tracked, using

individual months as samples (n = 13 months, t= -0.699, p = 0.498; Table 2.9).

94

Table 2.8. Monthly livestock use intensity values (LUVs) for radiolocations of

adult hyenas in the Talek West clan (95% Cl) collected during the entire month

(Overall Hyena), during only grazing hours (GH: 0900—1800 h) and livestock-free

hours (LFH: 1900—0800 h). For comparison, the average LUV for all cells in the

clan territory for each month (Territory) is also presented. Months with < 20

radiolocations in any time block were excluded. When the average LUV of hyena

locations is significantly greater than the territory LUV, “8” indicates selection for

livestock use areas relative to availability, and “A” indicates avoidance.

Livestock Use Values

 

 

Mo-Yr n (LFH, GH) Overall Hyena Territory LFH GH
Sep-02 125 (56,22) 10.2 (3.6) 9.4 11.9 (5.7) 10.3 (10.4)
Oct-02 126 (66,20) 6.9 (1.6) A 9.1 9.1 (2.6) 3.4 (1.7)
Dec-02 114 (53,25) 4.9 (1.7) 5.1 6.3 (1.5) 2.2 (1.6)
Feb-03 154 (67, 43) 1.4 (0.5) A 3.6 1.1 (0.2) 1.3 (0.6)
Mar-03 163 (97,24) 14.4 (3.5) S 7.9 17.0 (5.0) 13.7 (8.1)
Apr-03 110 (41,24) 5.2 (1.4) 4.0 5.0 (3.0) 7.7 (2.8)
Sep-03 116 (39,44) 17.0 (4.2) S 4.8 20.7 (6.7) 12.4 (5.1)
Oct-03 186 (94,23) 8.4 (2.2) 6.5 10.1 (3.1) 13.5 (8.6)
Nov-03 169 (87, 33) 2.7 (1.2) A 4.3 3.7 (2.2) 1.5 (0.3)
Dec-03 203 (90,47) 10.6 (1.5) S 5.2 9.2 (3.5) 9.3 (5.1)
Jan-04 160 (55,49) 2.7 (1.3) 2.4 3.1 (2.1) 2.3 (1.3)
Feb-04 201 (107,32) 6.0 (1.1) 6.4 6.6 (3.1) 6.8 (7.7)
Mar-04 216 (107,40) 6.0 (1.8) S 3.0 5.9 (1.8) 10.6 (7.7)

 

95

Table 2.9. Average livestock use intensity values (LUVs) for radiolocations of

adult hyenas in the Talek West clan (95% Cl) collected during grazing hours

(GH: 0900—1800 h) and during livestock-free hours (LFH: 1900-0800 h). LUVs

were derived from fixed-kernel grid surfaces based on all herds observed during

the monitoring period of each hyena. Hyenas monitored for < 18 months were

excluded. For comparison, the average LUV for all cells in the Reserve portion of

the clan territory, based on entire study period, was 12.6 (18.6). There was no

overall difference in average LUV of locations from GH and LFH using individual

hyena means as samples (paired t= -1.22; p = 0.25).

 

Hyena n Mean LUV - LFH n Mean LUV - GH
ALI 110 30.9 (4.5) 56 29.5 (6.9)
ATH 75 24.3 (5.1) 23 25.5 (11.0)
CSN 78 11.9 (3.3) 22 18.6 (6.5)
FOZ 66 24.6 (5.1) 25 26.2 (8.8)
LDV 81 21.3 (4.5) 30 29.1 (8.8)
MALI 146 24.9 (3.9) 67 24.5 (6.1)
MIG 60 16.9 (3.5) 30 20.1 (5.9)
MLN 112 20.3 (3.9) 46 20.3 (5.9)
MRPH 129 24.3 (3.3) 60 21.3 (4.9)
NICK 110 24.0 (3.7) 47 21.6 (6.3)
VGS 113 18.5 (3.7) 42 19.2 (6.9)
Overall Y 11 hyenas 22.0 (2.9) 11 hyenas 23.3 (2.4)

 

96

Table 2.10. Monthly average distances to, and % inside, the livestock core area
(LCA) for hyena radiolocations collected during grazing hours (GH: 0900—1800 h)
and livestock-free hours (LF H: 1900—0800 h). The LCA was the area of most
intense livestock use during the study period. Months with < 20 radiolocations in

any time block were excluded, except for the last row.

 

 

# of Locs. % in LCA Distance to LCA

Mo-Yr LFH, GH LFH GH f1 LFH GH t
Sep-02 (56, 22) 10.7 9.1 0.05 1520 1305 0.98
Oct-02 (66, 20) 9.1 0.0 1.95 1497 1610 -0.51
Dec-02 (53, 25) 7.5 0.0 1.99 1402 1777 -2.14*
Feb-03 (67, 43) 6.0 2.3 0.80 825 881 -0.38
Mar-03 (97, 24) 16.5 4.2 2.42 691 920 -1.00
Apr-03 (41, 24) 4.9 12.5 1.24 912 484 204*
Sep-03 (39, 44) 0.0 0.0 - 747 818 —0.56
Oct-03 (94, 23) 6.4 17.4 2.86 788 844 -o.32
Nov-03 (87, 33) 4.6 0.0 1.57 729 874 -1.29
Dec-03 (90, 47) 6.7 2.1 1.31 938 1004 -0.66
Jan-04 (55, 49) 10.9 6.1 0.75 970 1342 -2.09*
Feb-04 (107,32) 7.5 6.3 0.06 1475 1950 ~2.70*
Mar-04 (107, 40) 11.2 7.5 0.44 1390 837 3.05*
All Mo. (1277, 518) 10.1 9.5 0.17 1016 1030 0.32

 

' Significant test statistic (either f or t-statistic) at a = 0.05

97

DISCUSSION

Ecological variables influencing hyena space use

The first objective of this study was to identify the ecological factors
influencing space use in spotted hyenas. All the independent variables monitored
here were identified as important in predicting hyena space use patterns to some
extent, including vegetation types, water features, and the distribution of prey
resources. Given the social and biological importance of the communal den to
Crocuta, as discussed above, we expected den location to have a strong
influence on space use patterns. Model results reinforced this prediction with den
location selected as an important predictor variable in all models. This indicates
that the selection of den sites by females in this species may have serious
consequences for space use patterns of all clan members. Den sites selected by
wolves have been associated with avoidance of the territory edge (Ciucci & Mech
1992), villages, forest edges and intensively used roads (Theuerkauf et al.
2003c). It has also been suggested that wolf dens may be located in areas
providing a stable food supply (Ciucci 8 Mech 1992). In contrast, den selection
by Iberian lynx (Lynx pardinus) appeared unrelated to prey densities or
vegetation structure, but was closely associated with features of the microhabitat
at the den itself (Fernandez & Palomares 2000).

Den site selection in hyenas is poorly understood. Hyena den location was
often associated with water resources in both the Kalahari Gemsbok National
Park (Mills 1990) and the Mara Reserve (Boydston et al. 2006), and hyenas in

the Serengeti National Park typically selected dens located in the direction of

98

large prey herds in the dry season (Kruuk 1972). The data shown here lend
support to the idea that changes in location of the den site may be associated
with local changes in the spatial distribution of prey, as prey values were
negatively correlated with distance to the den during all but 1 of the 7 modeling
periods. This trend was particularly strong in models MR2 and TKW4, where
dens were clearly located in areas of high prey density (Figure 2.2 and 2.3).
Regardless of its influence on den site selection, the distribution of prey
clearly influenced space use decisions made by Reserve hyenas. Univariate
comparisons and multivariate models for 6 of the 7 modeling periods indicated
that prey numbers were significantly higher at “used” locations than at “available”
locations, demonstrating selection for prey-rich areas by hyenas in both clans.
Numerous studies have demonstrated a link between carnivore space use and
the prey densities within their home-ranges or territories (e.g., bobcat—Litvaitis et
al. 1986, caracal—Avenant & Nel 1998, lion—Hopcraft et al. 2005; Schaller
1972), and the same association has been shown in Crocuta. For example, in
southern Africa hyenas seasonally shifted movements within their territory to
utilize areas with the most abundant prey resources (Mills 1990; Trinkel et al.
2004), and extra-territorial movements increased as Serengeti hyenas sought out
distant migratory herds when local prey were scarce (Hofer & East 1993a).
Similarly, our data, and those of Boydston et al. (2003b), provide direct,
quantitative evidence that local movements of prey influence hyenas’ use of

space within their territories.

99

Water features were found to be important predictors of space use in both
clans. Although the location of water sources is suggested to influence
movements of hyenas in arid ecosystems (Cooper 1989; Tilson & Henschel
1986), it is unlikely that the hyenas studied here are limited by the presence of
water on the landscape. Because vegetation is clearly influenced by streams
and rivers in the Reserve (Figure 2.2), the importance of streams may simply
reflect the documented selection for shrubland vegetation. MR hyenas exhibited
stronger selection for streams than did TKW hyenas. This is likely a result of
selection for wet, muddy, concealed daytime resting sites in non-vegetated creek
beds, which were more common in the MR than TKW territory. This conclusion
is supported by the fact that selection for streams in MR was strongest during the

daylight hours.

The influence of livestock grazing on patterns of space use

The second objective of this study was to compare space use patterns of
hyenas between territories that differed in exposure to livestock grazing. With this
comparison, we investigated the broad hypothesis that livestock grazing was
influencing hyena space use decisions. Because Boydston et al. (2003b) found
increased use of vegetative cover over time by Talek hyenas associated with a
concurrent increase in livestock grazing within their territory, we predicted that
vegetative cover would assume relatively low importance to hyenas not exposed
to livestock grazing. Although hyenas in both clans demonstrated selection for

shrubland and avoidance of grassland relative to its availability, logistic

100

regression modeling confirmed our prediction, and indicated that vegetation class
was a more effective predictor of space use patterns in TKW than in MR.
Furthermore, models separated by time of day indicated that the selection by MR
hyenas for shrubland occurred almost exclusively during the day, yet selection
for shrubland in TKW was apparent regardless of time of day. Therefore, reduced
use of grass plains and increased use of shrubland habitat in the TKW relative to
MR clan appears to represent a behavioral modification resulting from the long-
term disturbance of grazing, as opposed to a daily response to the immediate
presence of livestock. These data support the suggestion by Boydston et al.
(2003b) that observed changes in use of vegetative cover over time in the Talek
clan resulted from increased exposure to livestock grazing.

Although den location was an important predictor of space use by
members of both clans, it was not consistently the most important predictor in
TKW models, and the reduced influence of the den on hyena locations in TKW
relative to MR is striking even on simple plots of tracking locations (Figure 2.8
and 2.9). This difference between clans was clear despite a generally higher
proportion of locations coming from females with den cubs in TKW.

The fact that vegetation class was a more important predictor variable
than den location in 2 of the 4 TKW models may indicate that TKW hyenas were
sacrificing proximity to the den for selection of vegetative cover. It is important to
note that in all but the TKW3 period, TKW hyenas chose to den along or near the
Talek River, which serves as the border of the Reserve in this area. Numerous

Maasai villages are located <1 km from this Reserve border (Kolowski &

101

Holekamp 2006). In addition, all livestock herds entering the Reserve must cross
this river, and often did so within 200 m of these den locations (J. M. Kolowski,
pers obs.). Although human disturbance has been associated with den
abandonment in coyotes (Bekoff & Wells 1982) and wolves (Ballard et al. 1987),
these hyenas seem to have selected particularly disturbed areas for their den
locations, and the reasons for this remain unknown. However, despite the fact
that TKW hyenas denned along this river for 17 of the 20 months of our study
period, the den exerted a strong influence on TKW space use in all modeling
periods, even during hours when livestock were present. Notably, in all cases,
these river dens were associated with dense shrubby vegetation and, in some
cases, with riparian forest habitat.

Because previous research indicated that areas heavily grazed by
livestock tended to have high prey densities (Boydston et al. 2003b), and
because any avoidance of livestock within a territory should result in sub-optimal
resource use, we also predicted that prey distribution would therefore be a more
effective predictor of hyena space use in livestock-free territories. In other words,
hyenas in livestock-free environments should make space use decisions based
on the resources critical to their survival and fitness, whereas those exposed to
disturbance may be forced to limit the cost of persecution at the expense of
optimal resource use. Our results supported this prediction. Although prey
distribution was important in predicting MR hyena space use in all modeling
periods, the influence of prey distribution on TKW hyena space use was

inconsistent, and varied in direction and importance among modeling periods and

102

between models based on locations collected during grazing hours and livestock-
free hours. Interestingly, the period during which hyenas exhibited avoidance of
areas with higher prey values (TKW1) was that during which livestock grazing
was observed to be most intense.

Based on data from our 4 modeling periods, we did not identify a
consistent spatial relationship between prey values and livestock use values.
The lack of a positive correlation between these variables indicated that the cost
to hyenas of avoiding livestock, in terms of loss of prey resources, was not
particularly high. Notably, although a strong positive correlation did exist
between these prey value and livestock use value in TKW2, prey distribution was
the most effective predictor of hyena space use for that period, indicating
selection for prey rich areas despite heavy use by livestock in these same areas.
Nevertheless, we found no consistent relationship between grazing intensity and
prey distribution among modeling periods.

Variables other than those monitored in this study (e.g. lion distribution
and tourist activity) may potentially influence space use patterns of hyenas in
other systems or of other carnivores. We did not take the distribution of either
lions or tourist vehicles into account in modeling the space use of Reserve
hyenas for the following reasons. Although lions are a major source of mortality
for Talek hyenas (Watts & Holekamp in review), Boydston et al. (2003b) showed
a positive spatial correlation between lions and hyenas in the Talek clan territory,
indicating space use by these carnivores is influenced by similar ecological

factors. In addition, we have shown previously that lion density did not differ

103

between the MR and TKW clan territories (Kolowski & Holekamp 2007).
Therefore, we were confident that lions were not likely to directly influence hyena
space use at the scale of interest here, or bias comparisons between the two
study clans.

We also did not incorporate tourist vehicle distribution into modeling
efforts. Tourist use of the Talek area was estimated to be almost five times
higher than that observed in the MR territory (Kolowski & Holekamp 2007).
However, tourist distribution is almost completely dependent on the distribution of
charismatic wildlife (e.g. cheetah, elephant, lion); a category in which hyenas are
generally not included. Therefore the local spatial distribution of tourist vehicles
is not predictable over time, making vehicle distribution mapping highly dubious.
Furthermore, hyenas have not been shown to avoid tourist vehicles in the
Reserve, even at close distances, due to frequent exposure and habituation (K.
E. Holekamp, unpublished data).

Numerous studies have shown carnivores to utilize spatio-temporal
avoidance to accommodate human activity or disturbance in their territories. For
example, grizzly bears have been shown to increase their use of areas near
roads at night, when vehicular traffic was relatively light (Gibeau et al. 2002;
McLellan & Shackleton 1988). Wolves in Italy used paved roads, villages and
rubbish dumps almost exclusively at night (Ciucci et al. 1997). And in British
Columbia, portions of rivers heavily used by anglers received relatively more

night use from grizzly bears than portions with fewer anglers (Machutchon et al.

104

1998). Even in this hyena population, we found that use of areas around Maasai
villages occurred almost exclusively at night (Kolowski & Holekamp 2006).

Given the temporal predictability of livestock grazing in the Reserve, and
the direct threat that herders pose to hyenas, we expected areas used heavily by
livestock to be used to a greater extent by hyenas during the night, when this
disturbance was absent. However we found no evidence of this. We
documented little avoidance of livestock use areas in general, and no consistent
differences in the use of these areas based on time of day. TKW hyenas appear
to have adjusted to the presence of livestock within their territory, and their
increased use of vegetative cover apparently allows hyenas to continue to make
space use decisions based on den location and prey distribution, regardless of
time of day, or livestock distribution. Although reduced daytime activity by TKW
relative to MR hyenas (Kolowski et al. 2007) may have reduced the need for
direct avoidance by TKW hyenas of areas used by livestock, our data suggest
the availability of refuges during daytime rest periods is important in allowing
shared use of space.

However, we found some evidence suggesting that there may be a
threshold level of grazing intensity, beyond which new patterns of space use may
be required. Livestock use values were unimportant in predicting hyena space
use patterns in all logistic regression models except TKW1, when livestock
grazing was most intense, and when higher livestock use values were associated
with a lower probability of hyena use. During December, February and March of

this period, the % of locations in the livestock core area (LCA) was always higher

105

during livestock-free hours (LFH) than during grazing hours (GH), and locations
during LF H were on average closer to the LCA than those collected during GH
(though neither trend was significant). Thus it may be that, at the levels generally
observed during this study (except during TKW1), direct avoidance of livestock
was not necessary, and increased use of vegetative cover was adequate to

accommodate the disturbance.

Conclusions

The data presented in this chapter lead to two primary conclusions. First,
and most broadly, patterns of space use in this species are the result of a
combination of multiple ecological factors, including prey distribution, vegetation
types, water features on the landscape, and is most strongly influenced by the
location of the communal den; a complete understanding of space use in this
species requires a more comprehensive understanding of the factors influencing
den site selection.

Second, given the relationship between the distribution of livestock,
vegetation and hyena space use, we can also conclude that the presence of
vegetative cover appears to be critical in allowing the coexistence of livestock
and hyenas at a small spatial scale. We demonstrated that daily, often intense
livestock grazing pressure did not result in measurable spatial avoidance of
grazed areas and that despite this disturbance, TKW hyenas were largely able to
modify patterns of space use to maximize proximity to the communal den and to

the areas within their territory that contained the highest prey densities. We

106

hypothesize that in the absence of vegetative cover this coexistence would not
be possible, and that spatial avoidance of intensely grazed areas, either during
grazing hours or at all times of day, would be dramatic. This is supported by the
fact that the spatial avoidance of intensely grazed areas shown by Boydston et
al. (2003b) occurred in areas of open plain containing virtually no dense
vegetative cover.

Documentation of behavioral responses to disturbance assumes its
greatest utility to conservation when it is linked with demographic consequences
(Caro 1998; Gill & Sutherland 2000). A detailed investigation into the
demography of the Talek Clan from 1988 to 2003 indicated that the population
was stable, and birthrates did not decline over this time period. This was despite
the fact that humans were a significant source of mortality for adult hyenas in this
clan, and that the annual mortality rate due to humans was more than 4 times
greater from 1996—2003, than from 1988—1995 (Watts & Holekamp in review).
Given these data and the results presented above, it appears that hyenas living
along this Reserve border have adjusted to coexist with the daily disturbance of
livestock grazing, and importantly, have not yet suffered measurable
demographic consequences due to either this disturbance, or the increased
source of mortality associated with their proximity to the Reserve border. It is
unlikely that growth of the local human population will result in a dramatically
increased livestock population, given that livestock numbers have remained
relatively stable over the last 20-30 years, and that current per capita stock

numbers are thought to be unsustainable (Lamprey 8 Reid 2004; Serneels &

107

Lambin 2001). However, a combination of degradation of rangeland outside the
Reserve (Serneels et al. 2001) and the increased threat of direct hyena mortality
associated with a growing human population, suggests that grazing pressure in
the Reserve and human-caused mortality will continue to increase over time.
Whether behavioral changes, such as increased used of vegetation (Boydston et
al. 2003b), and this study) and changes in activity patterns (Kolowski 8
Holekamp 2007), or demographic resiliency (Watts 8 Holekamp in review) will
continue to buffer these hyenas from these increasing threats remains to be

seen.

108

CHAPTER THREE

Kolowski, J.M., and K. E. Holekamp. In review. Effects of an open refuse pit on
space use patterns of spotted hyenas. African Journal of Ecology.

109

CHAPTER THREE

EFFECTS OF AN OPEN REFUSE PIT ON SPACE USE PATTERNS OF
SPOTTED HYENAS

INTRODUCTION

Patterns of space use by large carnivores are determined by a multitude
of factors including availability of water, den sites, cover from predation or climate
(Ewer 1973) and competition (Durant 1998), but are often most closely
associated with the distribution and abundance of prey (Hofer 8 East 1993b;
Mills 8 Knowlton 1991; Spong 2002). Numerous species utilize human-provided
food sources as components of their diet (e.g., Canis lupus—Fuller 8 Keith 1980;
Vulpes vulpes—Doncaster et al. 1990; Ursus amen'oanus—Herrero 1983; U.
arctos—Knight 8 Eberhardt 1985; Crocuta crocuta—Mills 8 Hofer 1998). It has
been shown these anthropogenic resources influence carnivore movements
(Ciucci et al. 1997; Craighead 8 Craighead 1971), home range size (Blanchard 8
Knight 1991; Hidalgo-Mihart et al. 2004), and population density (Fedriani et al.
2001; Fuller 8 Keith 1980). Additionally, the presence of these resources may
influence the frequency of human-carnivore conflict by increasing carnivore
densities (Yom-Tov et al. 1995), or by shifting populations (Beckmann 8 Berger
2003b) or individuals (Knight et al. 1988; Lunn 8 Stirling 1985) toward human-
dominated areas. Resulting conflicts may cause property damage (e.g.
livestock), human injury, and even increased carnivore mortality, with these
areas acting as population sinks (Knight et al. 1988). As human population
growth and habitat conversion proceed at increasing rates, human-carnivore

conflict, particularly livestock depredation, threatens to impede large carnivore

110

conservation efforts where carnivores share the landscape with humans.
However, few studies have investigated the influence of human-provided food
sources other than livestock on the movements and space use of potential
livestock predators.

Seasonal variation in the use of refuse sites has been documented in a
number of carnivores and is often associated with variation in prey availability
and seasonal variation in nutritional requirements (Craighead 8 Craighead 1971;
Lucherini 8 Crema 1994; Salvador 8 Abad 1987). It is also likely that variable
nutritional demands of different age/sex classes, and reduced competitive ability
in social carnivores of low social rank, may result in variation in the use of these
resources at the level of the individual. Identification of the animals most likely to
utilize human-provided food sources may help to identify the existence of
potential “problem individuals” (see review in Linnell et al. 1999), and the factors
that may lead to their destructive behavior. However, few studies have
investigated individual variation in the use of refuse sites by large carnivores (but
see Lunn 8 Stirling (1985) and Rogers et al. (1976)).

Spotted hyenas (Crocuta crocuta Erxleben) are important predators of
livestock throughout East Africa (Kolowski 8 Holekamp 2006; Kruuk 1981;
Ogada et al. 2003). Although Crocuta feed primarily on freshly killed ungulates
that they hunt themselves (Gasaway et al. 1991; Kruuk 1972), hyenas are
opportunistic foragers and efficient scavengers that have been observed to feed
on a huge variety of food items including insects, birds, rodents and even

garbage (Cooper et al. 1999; Henschel 8 Skinner 1990; Mills 8 Hofer 1998). It

111

has been suggested that refuse and livestock carcasses at pastoral villages may
influence the frequency of hyena visits to these areas and ultimately, livestock
depredation rates (Kolowski 8 Holekamp 2006; Mills 8 Hofer 1998).

Our goal here was to document the influence of a single human refuse
source on the space use patterns of adult hyenas, and to investigate the
influence of hyena social rank and local prey abundance on variation in the use
of this food source. We capitalized on a natural experiment in which a refuse site
was closed midway through intensive monitoring of individual hyenas whose

defended group territory contained the refuse site.

METHODS

Study population and habitat characteristics

Our study was conducted from June 2001 through April 2004 in the Masai
Mara National Reserve (hereafter the Reserve) in southwestern Kenya. The
1500 km2 Reserve consists primarily of rolling grassland and scattered bushland
(predominantly Croton and Euclea species), with riparian forest along the major
watercourses, and supports a large diversity of resident ungulates. From August
to October, the Reserve also supports large migratory herds of wildebeest
(Connochaetes taun'nus) and zebra (Equus burchelli). Due primarily to its
diversity and abundance of predators, as well as the seasonal influx of migratory
ungulates, the Reserve is Kenya’s premier wildlife tourist attraction (Norton-
Griffiths 1995), and supports 24 permanent tourist camps and lodges (Walpole et

al. 2003). Although many are located just outside the Reserve boundary, a small

112

subset of lodges (~20°/o) is located within the Reserve itself, and this study
focuses on one of the latter.

Spotted hyenas live in social groups called clans, and clan members
cooperatively defend a stable group territory. Each clan contains one to several
matrilines of adult females and their offspring, as well as a variable number of
adult immigrant males. Clans are rigidly structured by hierarchical rank
relationships (Frank 1986b; Kruuk 1972; Tilson 8 Hamilton 1984) that determine
priority of access to food, and all adult females are socially dominant to
immigrant males (Kruuk 1972; Smale et al. 1993). Subadult individuals of both
sexes maintain their maternal ranks as long as they remain in the natal clan
(Smale et al. 1993). Although females are generally philopatric (Frank 1986b),
virtually all natal males disperse between the ages of 2 and 5 years (East 8
Hofer 2001; Henschel 8 Skinner 1987; Smale et al. 1997). Crocuta clans are
fission-fusion societies in which individuals travel, rest and forage in subgroups
that typically change in composition many times each day (Holekamp et al.
1997a; Kruuk 1972). Female Crocuta bear 1—2 (rarely 3) young in isolated natal
dens (Holekamp et al. 1996). Cubs are typically transferred to a communal den
at 2-5 weeks of age where they reside for the next 7-8 months (Kruuk 1972).
The communal den represents the social center of each clan’s territory and most
clan members visit it regularly.

We monitored individuals from a single clan (the Mara River clan) that
defended a territory (31 km2) near the center of the Reserve (Figure 3.1). The

clan included 32—43 individuals (7 = 8 adult females, 3? = 5 adult immigrant

113

males), and each hyena was individually recognized by unique spot patterns. A
single tourist lodge, which maintained an unfenced refuse pit approximately 300
m from the periphery of its grounds, was located near the northern boundary of
the Mara River clan territory (Figure 3.1). Garbage, mainly composed of food
refuse from tourist and staff dining halls, was deposited daily into an earthen pit,
usually between 0900—1000 h. The pit was closed on 12 October 2002 and all
garbage was removed from the site. In the following weeks the pit was filled with

soil, and native shrubs and trees were planted.

Data collection

We anesthetized and radiocollared 6 adult female hyenas from the Mara
River clan of variable social rank, and made attempts to locate each individual on
a daily basis. On average, we collected 15 telemetry locations per month for
each collared hyena. The majority of radiotracking was conducted during the
morning (0530—0900 h) and evening (1730—2000 h), but additional locations
were collected during mid-day and throughout the night. In addition to tracking
efforts, we made regular visits to the refuse pit at all times of day; however,
hyenas were only observed at the site when fresh refuse was present. During
these observation sessions we recorded the number and identity of hyenas

present, as well as the presence of other species at the site.

114

 

      
 

  

Masai Mara
National Reserve

__ 8168‘ '

5' Jal‘cb

 

 

 

     
 

......

Mara River
Territory

.......

1 0 1 IGIometers
E

Figure 3.1. Territory of the Mara River hyena clan and its location within the
Masai Mara National Reserve. Hatched lines represent ungulate prey sampling

transects. The refuse pit (filled circle) and tourist lodge are also indicated.

115

Using tracking locations from 5 of the 6 monitored adult female hyenas
(one female was not monitored after pit closure), we compared space use
patterns before and after pit closure using the following variables: 95% fixed
kernel home range size, 50% fixed kernel home range (core area) size,
presence/absence of the refuse pit within the core area, and average hyena
distance to the refuse pit. Home range size during the full 16-month pre-closure
period was strongly influenced by the consecutive use of communal dens
separated from each other by as much as 6 km. We therefore limited the pre-
closure locations to those collected in the 10 months prior to pit closure, when
consecutive den locations were consistently only short distances apart (this held
for the remainder of the study period). Because space use of female hyenas is
dependent on whether they have cubs residing at the communal den (Boydston
et al. 2003a), we categorized all pre-closure locations for each female with
respect to whether or not she had den-dwelling cubs. We then randomly
subsampled locations collected in the 19-month post-closure period for each
female to equalize the number of locations collected before and after closure with
and without den-dwelling cubs (minimum of 20 locations per female pre- and
post-closure). Because female space use is even further restricted during use of
natal dens, we excluded locations collected at these dens from all analyses.
Using this restricted dataset, we calculated a single 95% and 50% home range
(with smoothing factors determined by least squares cross-validation—Seaman
8 Powell (1996)), and an average distance at which each female was tracked

from the refuse pit, for the pre- and post-closure periods.

116

The pooling of locations to calculate home range and core area size, while
necessary due to sample size requirements, prevented statistical comparisons of
these variables before and after pit closure. We compared average distances to
the refuse pit before and after closure using a Wilcoxon matched-pairs test. All
distances and fixed kernel home ranges were calculated using Animal Movement
Analyst (Hooge 8 Eichenlaub 2000) and ArcView GIS 3.2 (Environmental
Systems Research Institute, Redlands, California).

Because variability in prey abundance over time may influence space use
patterns, we monitored prey abundance within the clan territory throughout the
study period. As a monthly index of local prey abundance, we counted the total
number of wild ungulates within 100 m of 2 4-km transects twice each month
(Figure 3.1), and calculated the average number of prey animals counted per
sampling event. Prey abundance was estimated based on 10 months of
sampling data during the pre-closure period and compared, using a t-test, to an
estimate based on the same 10 months during the post-closure period to control
for the seasonal ungulate migration.

To investigate the influence of social rank on individual variation in space
use relative to the refuse pit, we conducted two separate analyses with females
assigned to low or high rank categories relative to the median adult female rank.
Individual ranks were assigned based on the outcomes of dyadic agonistic
interactions as in (Holekamp 8 Smale 1990). Immigrant males were treated
separately as the lowest ranking group. First, to qualitatively compare use of the

pit between individuals of different ranks, we calculated the proportion of tracking

117

locations collected within 500m of the refuse pit for those adult females tracked
for 212 months while the pit was in use. Second, independent of radiotracking,
we documented the composition of hyena groups observed feeding at the refuse
pit based on 18 observation sessions. We compared the proportion of these
sessions recording the presence of low-ranking females, high-ranking females
and immigrant males. We then compared the number of hyenas present from
each rank group at these visits using a Friedman ANOVA, with the number of
hyenas from each rank group representing three repeated samples for each
observation session. Each rank group contained a total of five individuals in this
pre-closure period.

To investigate the influence of natural prey abundance on use of the
refuse pit, we compared pit utilization during months of high prey abundance,
when migratory herds were present in the clan territory, to months when only
resident herds were present. Migration months were defined as those months
when wildebeest were seen within the territory as no resident herds of wildebeest
utilize this portion of the Reserve. We focused only on hyenas that were closely
monitored in the pre-closure period (212 months of tracking), and that were
known to use the site with some frequency (>5% of locations within 500 m of the
refuse pit). We compared the relative frequency of locations of these hyenas
(pooled) within 500 m of the refuse pit during migration months and non-

migration months using a chi-square test.

118

All tests were considered statistically significant at on = 0.05 (two-sided),
and all analyses were conducted using the STATISTICA software package

(StatSoft 2002). Descriptive statistics are presented as means 1- SE throughout.

RESULTS

Observations of refuse pit use

At least one clan member was present at 18 visits by researchers to the
refuse pit. During these sessions an average of 5 hyenas were seen at the site
(range: 2—14). The hyenas were often joined at the refuse pit by savanna
baboons (Papio cynocephalus), warthogs (Phacochoerus afn'canus) and various
species of vultures. Most edible items were typically consumed within 1—2 hrs of
refuse deposition. Typically only a subset of the hyenas present, if any, was
observed to feed, due to direct and sometimes aggressive competition with
warthogs and baboons. At least one subadult hyena was present at 44% of
these sessions, the youngest of which was approximately 13 months. None of
the subadults were accompanied by their mothers and no hyenas <1 year old

were observed at the refuse pit.

Effect of refuse pit on clan space use

We collected a total of 1830 locations from 6 radiocollared females, with
each female monitored for an average of 20 months. Based on locations from
the 5 females monitored before and after pit closure, home range in the 10

months prior to closure was 10.9 km2 (n = 385 locations), and 13.9 km2 (n = 385

119

locations) in the 19 months after closure (Figure 3.2). The pre-closure core area
was 1.6 km2 and included the refuse pit as well as 2 of the 6 communal dens
utilized during this period (Figure 3.2A). After pit closure the core area no longer
included the pit, was almost half its previous size (0.9 kmz), and contained 7 of
the 11 communal dens utilized during the post-closure period (Figure 3.2B).
Distance to the pit was significantly smaller during the pre-closure period (7 =
1.87 i 0.09 km) compared to that in the post-closure period (Y = 2.23 :t 0.07 km;
Z = 2.00, p < 0.043; Table 3.1). Differences in space use between the pre- and
post-closure periods could not be explained by differences in prey abundance.
Average monthly prey counts were similar in the pre— and post-closure periods

(pre: 3? = 294, post: 7 = 394; t= -0.522, p = 0.607).

Individual variation in refuse pit utilization

Of 6 radiocollared females, four (2 high-ranking and 2 low-ranking) were
monitored for 312 months during operation of the refuse pit. The two high-
ranking females were tracked within 500m of the pit on only 2 of 179 (1.1%) and
4 of 202 (2.0%) occasions, respectively. In contrast 29.8% of 178 locations, and
26.2% of 183 locations for the two low-ranking females were within 500m of the
pit. Seventy-five percent of 18 refuse pit observation sessions recorded low-
ranking adult females present, whereas only 43% and 32% recorded the
presence of immigrant males and high-ranking females, respectively. At the

same 18 sessions these three rank groups were not represented equally

120

Table 3.1. Average distances (SE) of adult female hyena tracking locations to
the refuse pit for the 10 months prior to the closing of the pit (pre-closure) and the
19-months following the closure (post-closure). The rank group of each hyena is

indicated after each name (H=High, L=Low).

Mean Distance (km) to Refuse Pit

 

 

Hyena # of Tracks per Period Pre-Closure Post-Closure
NANA(H) 130 2.19 (0.14) 2.51 (0.13)
ATAR(L) 119 1.89 (0.18) 1.98 (0.15)
BACK(L) 96 1.58 (0.18) 2.07 (0.13)
WND(L) 21 1.57 (0.38) 2.62 (0.22)
CHAC(L) 20 1.39 (0.19) 2.25 (0.32)
Pooled Locations 385 1.87 (0.09) 2.23 (0.07)

121

 

2 Kilometers

Figure 3.2. The 95% home ranges (solid dark lines) and 50% core areas (dotted
lines) based on locations from 5 radiocollared adult female hyenas collected
during the 10 months before (A) and 19 months after (B) closure of the refuse pit.
Pit location is indicated as in Figure 3.1, and the locations of all communal dens

utilized during the respective periods are indicated by solid triangles.

122

(Friedman ANOVA X2 = 15.5, p < 0.0005), with low-ranking females being the

most numerous individuals at the refuse pit (Figure 3.3).

Influence of natural prey abundance on refuse pit utilization

Migratory herds were present in the study area for 3 of the 9 months for
which both prey and tracking data were available during operation of the refuse
site. Unfortunately, only two females in the clan met our criteria for this analysis,
yet clear trends were shown by these individuals. These two females were
tracked to within 500 m of the pit significantly less often during migration months
(12.0% of all tracks) than during non-migration months (38.0% of tracks; X2 =

19.24, p < 0.0001).

DISCUSSION

In carnivore studies, utilization of human-provided food resources such as
those available at garbage dumps often results in reduced individual or group
home ranges (e.g. C. latrans—Hidalgo-Mihart et al. 2004, U. arctos—Blanchard
8 Knight 1991) or core areas (e.g. Mungos mango—Gilchrist 8 Otali 2002).
These observed reductions are presumed to result from a reduction in foraging
space requirements due to the addition of a predictable, concentrated food
supply . Here, the female group home range was smaller during operation of the
refuse pit, yet core area size was smaller following closure of the pit. This
discrepancy is probably related to the importance of the communal den in hyena

society. Although its influence on space use varies with rank and reproductive

123

2.2
2.0
1.8
1.6 L
1.4+ ——
1.2-

j'

 

1.0l ——

0.6 ~
0.4 - u
0.2 -
0.0

Mean # of Hyenas Present at Pit

 

 

#—

 

 

HR Female LR Female Immigrant Male

Figure 3.3. Mean number of high-ranking (HR) females, low-ranking (LR)
females, and immigrant male hyenas (5 individuals per group) that were seen at
18 refuse pit observation sessions. Sessions included were those at which at

least one hyena was observed. Whiskers indicate 95% confidence intervals.

124

condition, space use by all female clan members is highly influenced by the
location of the communal den (Boydston et al. 2003a). It appears that during its
operation, the refuse pit became a focal point of activity for at least some clan
members, resulting in an expanded group core area that contained communal
den locations and the refuse site. Unlike in garbage-feeding banded mongoose
and baboon groups, where core areas were centered on refuse site locations
(Altmann 8 Muruthi 1988; Gilchrist 8 Otali 2002), the location of the communal
den likely dictates the location of hyena clan core areas. And, further, the
location of the den appeared to be independent of the pit location, because den
location was similar between the pre— and post-closure periods (Figure 3.1).

Use of the refuse pit was most common among low-ranking females,
particularly during times of prey scarcity, though there was notable variation in pit
utilization within rank groups. For example, at least one low-ranking uncollared
female was seen only once at the refuse pit, and an uncollared high-ranking
female was observed at the pit with some regularity. Based solely on rank and
its associated priority of access to food, we expected immigrant males to be most
dependent on alternate food resources, yet immigrant males were less common
at the refuse site than low-ranking females. This likely resulted from exclusion of
immigrant males at the pit by the higher-ranking female hyenas. Notably,
frequent refuse pit users in our study were not infirm or consistently old, and all
were known to be capable of making kills and consuming ungulate carcasses.

The observed shifts in space use, and the propensity of certain clan

members to frequent the refuse pit are particularly interesting as they relate to

125

human-carnivore conflict. We suggest that the relatively frequent use of the
refuse pit by low-ranking individuals is related to their low priority of access to
kills in the clan territory. This skewed utilization of human food sources has been
similarly shown in polar bears (Lunn 8 Stirling 1985) and black bears (Rogers et
al. 1976; Young 8 Ruff 1982), where subadults (particularly males in black bears)
were the most common age group at garbage dumps. This has been suggested
to result from the increased nutritional stress (Lunn 8 Stirling 1985), ranging
behavior (Rogers et al. 1976) or necessity of avoiding intraspecific competition
(Young 8 Ruff 1982).

If human refuse attracts hyenas to pastoral villages then, given our results
here, we would expect low-ranking hyenas to be more likely to visit these villages
and perhaps also to opportunistically attack corralled livestock there. From
2001—2005, 11 known hyenas from two clans defending territories along the
Reserve border were found with neck snares, often set at livestock enclosures.
Seventy-three percent of these were either immigrant males or low-ranking natal
hyenas. During the same period, six of the seven hyenas from these clans that
were either killed during livestock depredation attempts or found dead near
villages were also low-ranking. While the possibility exists that low-ranking
hyenas are more likely to seek out and attack livestock directly, our space use
data suggest that garbage alone can influence movements of hyenas, particularly
those of low rank, and that these individuals may therefore be more likely to

participate in depredation events.

126

The potential of the presence of human-provided food sources to influence
carnivore space use patterns is well documented. A wolf pack in Italy focused
their nightly movements around garbage dumps (Ciucci et al. 1997), and there is
a long history in North America of both black bears (Herrero 1983; Rogers et al.
1976) and grizzly bears (Craighead 8 Craighead 1971) visiting garbage dumps
with remarkable frequency and predictability. Beckmann and Berger (2003) have
related use of these dumps by black bears to an increase in direct conflict with
humans, but this problem may be exacerbated when sources of garbage occur in
the vicinity of livestock. In an earlier study we showed that larger pastoral
villages along the Reserve border were more likely to suffer livestock losses to
hyenas, and suggested that large villages may be more attractive to foraging
hyenas due to the larger amounts of refuse produced by them (Kolowski 8
Holekamp 2006). We have shown here that even small sources of human refuse
have the ability to alter hyena space use patterns and result in disproportionately
frequent use of these areas. As refuse and livestock carcasses are common
around many pastoral villages surrounding the Reserve, it seems likely that
regular nightly movements into village areas by Reserve hyenas (Kolowski 8
Holekamp 2006) are associated with utilization of this resource. Because
increased hyena use of these village areas is associated with more frequent
hyena attacks on livestock (Kolowski 8 Holekamp 2006), the ability of human
refuse to influence hyena movements makes the presence of these food sources
in and around pastoral villages a potentially serious obstacle to efforts to reduce

the frequency of livestock depredation events.

127

CHAPTER FOUR

Kolowski, J.M., and K. E. Holekamp. 2006. Spatial, temporal, and physical
characteristics of livestock depredations by large carnivores along a Kenyan
reserve border. Biological Conservation 128:529-541.

128

CHAPTER FOUR

SPATIAL, TEMPORAL, AND PHYSICAL CHARACTERISTICS OF LIVESTOCK

DEPREDATIONS BY LARGE CARNIVORES ALONG A KENYAN RESERVE

BORDER
INTRODUCTION
Human activity has caused a global decline in many large carnivore
species (Fuller 1995; Nowell 8 Jackson 1996). As a result, large carnivores have
disappeared from areas of high human density, and the current risk of extinction
to a given predator species is closely linked with its level of exposure to human
populations (Woodroffe 2001). Although habitat conversion, declining natural
prey populations, and commercial exploitation have contributed to carnivore
losses, active persecution by humans, based on real or perceived threats to
themselves and their livestock, appears to be the most important factor in
observed declines (Woodroffe 2001). Large carnivores, humans and their
livestock have coexisted for millennia, but recent decades have seen dramatic
increases in the frequency of human-carnivore conflict, resulting mainly from an
exponential increase in the human population (Conover 2002; Woodroffe 2000).
In the face of such persecution, protected areas are fast becoming the last

refugia for many large African predators (Mills 1991), all of which have
experienced significant declines in recent decades (Ginsberg 8 Macdonald 1990;
Mills 8 Hofer 1998; Nowell 8 Jackson 1996; Ogutu et al. 2005). However, even
within protected areas, humans often remain the main source of mortality to large

carnivores (Woodroffe 8 Ginsberg 1998), with smaller reserves surrounded by

dense human populations being particularly susceptible to species loss

129

(Brashares et al. 2001; Harcourt et al. 2001). Given current rates of habitat
conversion and human population growth, expansion of African reserves is
unlikely, yet few of Africa’s existing reserves are large enough to maintain viable
populations of wide-ranging predators (Brashares et al. 2001). Therefore,
conservation of large African carnivores is likely to depend on networks of
smaller reserves, buffer zones, and private and communal lands, where
successful conservation will be closely linked with an ability to resolve human-
carnivore conflicts and minimize numbers of carnivores killed by people
(Woodroffe 2001). To succeed in these efforts, park managers, biologists, and
indigenous people must coordinate efforts to understand the circumstances
surrounding conflicts involving large carnivores, and combine systematic data
with local experience to identify factors that may mitigate conflict (Treves 8
Karanth 2003).

The Masai Mara National Reserve, Kenya (hereafter the Reserve) is one
of East Africa’s most popular game viewing locations, largely because it supports
a high density of large carnivores (Ogutu 8 Dublin 1998; Ogutu 8 Dublin 2002).
With the passage of the Land Group Representatives and Land Adjudication Act
of 1968, much of the communal land immediately north of the Reserve was
opened to demarcation into group ranches to formalize security of land tenure for
the predominantly pastoral resident Maasai people (Kimani 8 Pickard 1998). By
the late 1970s all rangelands surrounding the Reserve had been assigned to
group ranches, 4 of which currently surround the Reserve: Lemek, OI Kinyie,

Koyake and Siana. The rangeland within these group ranches was intended to

130

act as a buffer zone, protecting known wildlife dispersal areas from habitat
conversion, and separating the protected area within the Reserve from
expanding commercial agriculture to the north (Serneels et al. 2001). However,
as available rangeland north of the ranches is lost to agriculture and the human
population continues to grow (by an estimated 4.4% per annum on the Koyake
group ranch (Lamprey 8 Reid 2004)), carnivore-livestock conflict is inevitable and
likely to increase. To avoid the establishment of population “sinks" surrounding
the Reserve, in which human-caused mortality limits survival of predators
dispersing from the Reserve (Woodroffe 8 Ginsberg 1998), livestock depredation
and the resulting persecution of carnivores must be minimized.

Rates of livestock depredation by large carnivores can be influenced by
local environmental conditions such as abundance of natural prey (Meriggi 8
Lovari 1996; Mizutani 1999; Polisar et al. 2003; Stoddart et al. 2001) and rainfall
(Patterson et al. 2004; Woodroffe 8 Frank 2005), as well as socio-ecological
factors including livestock husbandry practices (Ciucci 8 Boitani 1998;
Madhusudan 2003; Meriggi 8 Lovari 1996; Ogada et al. 2003; Stahl et al. 2001)
and characteristics of attacked farms, villages, and livestock enclosures (Mech et
al. 2000; Ogada et al. 2003). However, few studies have concurrently
investigated the influence of both environmental and socio-ecological factors on
livestock depredation, and even fewer have combined this knowledge with
consideration of the behavior and movements of monitored predators. Our goal
was to elucidate relationships between various ecological factors and temporal

variation in conflict frequency in the vicinity of the Reserve, and to assess the

131

influence of village and enclosure characteristics on relative vulnerability to
carnivore attack. Finally, the concurrent long-term study of spotted hyenas
(Crocuta crocuta) in our study region provided a unique opportunity to associate

detailed data on hyena movements with hyena depredation behavior.

METHODS

Study area
Geography and Climate

Our study was conducted along the northeastern border of the Masai
Mara National Reserve (1500km2) in southwestern Kenya (Figure 4.1). The
portion of the study area outside the Reserve was situated in sections of the
Koyake and Siana group ranches, which share their southern borders with the
northern border of the Reserve. The study focused on three adjacent locales
within these ranches: Talek, Ntipilikwani, and Olosogon. These were chosen for
their accessibility, and their relatively high density of settlements near the
Reserve. The Talek region supports the highest density of settlements along the
entire northern border of the Reserve (Reid et al. 2003). The dominant land uses
on the group ranches are subsistence pastoralism and wildlife tourism. Most of
the annual rainfall in this region falls during one of two wet seasons: the “short
rains” in November-December, and the “long rains” in March—May. During 2003,

total rainfall in our study area was 1223 mm.

132

 

Koyake GR

Masai Mara

National Rem e 1

’ Koyake GR

_, l Siana GR

4'5”
.1

 

 

 

 

1 0 1 2 Kilometers

Figure 4.1. Locations of Maasai bomas (villages) of the Talek (filled circles),
Ntipilikwani (open circles), and Olosogon (triangles) locales from which
information on livestock depredation by large carnivores was collected from
March 2003 to April 2004. Only bomas that contained livestock and were located
within 2 km of the Reserve border are included. Dashed lines indicate prey

transects and the town of Talek is indicated with a star.

133

Wildlife and Habitat

Both the Reserve and the surrounding group ranches support a large
diversity of resident ungulates including gazelles (Gaze/la thomsonii and G.
grantl), impala (Aepyceros melampus), topi (Damaliscus lunatus), and giraffe
(Giraffa camelopardalis). From August to October small resident populations of
wildebeest (Connochaetes taun'nus) and zebra (Equus burchelli) are joined by
large migrant herds from Tanzania. It is estimated that 300,000—750,000
wildebeest enter the Reserve each year during the migration, with 50,000—
150,000 spilling onto the four adjacent group ranches (Broten 8 Said 1995;
Ottichilo et al. 2000; Serneels 8 Lambin 2001). The dominant predators in the
Reserve are spotted hyenas, lions (Panthera Ieo), leopards (P. pardus) and
Cheetahs (Acinonyxjubatus). Although striped hyenas (Hyaena hyaena) also
occur in this region, they are rare and were not known to attack any livestock
during the study period. Therefore, discussions of hyenas below refer
exclusively to spotted hyenas.

The Reserve consists primarily of rolling grassland habitat and scattered
bushland (predominantly Croton and Euclea species), with riparian forest along
the major watercourses. The Koyake and Siana group ranch property in our
study area is grazed year-round by livestock, and includes habitat similar to that
of the Reserve, with somewhat reduced woody vegetative cover. The Talek
River and its tributaries support the majority of the vegetative cover in the study

area.

134

The Maasai Village

The traditional Maasai village, or boma, in this region consists of a
collection of huts, constructed from wooden frames covered with mud and dung,
surrounding a central cattle enclosure (Figure 4.2). Although sometimes used to
mean “fence”, the term boma here will refer to an entire village or settlement,
which may include numerous households and livestock enclosures (Burnsilver et
al. 2003; Homewood 8 Rodgers 1991). Typically a number of household heads
reside at a boma with their personal dwellings built in distinct sections of the
boma. Each household head keeps his cattle in the shared central enclosure at
night and maintains a separate personal enclosure among his huts, in which his
own sheep and goats are kept at night. Of the 78 bomas identified within our
study area, 64 (82.1%) of these contained a single cattle enclosure with < 2
sheep/goat enclosures. The remaining bomas contained one cattle enclosure
and, on average, three separate sheep/goat enclosures, roughly approximating
the number of resident livestock owners.

In our study area, livestock enclosures were constructed of local bush
(often thorned), or tall (1 .5—2 m) pieces of split timber (referred to as poles),
spaced up to 0.25 m apart, and sometimes fortified by chain link or barbed wire.
Enclosures for sheep and goats were always more sturdy and complete than
those for cattle. A small peripheral and often poorly maintained fence of bush
was often constructed around the entire boma compound (Figure 4.2). However,

this peripheral fence was always fortified where it comprised part of an individual

135

~ 0
" § Stock-owner A’s cattle

Centnl corral i (a hill“)
5,
2r
1‘

Stock-owner B‘s
homestead 111mm pens for small stock
(calves or goats/sheep at night)

 

Figure 4.2. Configuration of a typical Maasai village (boma). Cattle from all
household heads are housed together in the shared central corral. Individual

homesteads maintain separate enclosures for their own small stock. Figure

adapted with permission from Spencer (2003) pg 45.

136

livestock enclosure. Donkeys were rare in the region, but when present were
usually kept at night within the peripheral boma fence.

Livestock were typically driven out of the boma between 0800-0900 h for
grazing, and returned to the boma just before sunset. All herds outside of the
boma are referred to here as grazing herds. Cattle typically traveled 2—5 km
from the boma during the day, but sheep and goat herds generally remained < 2
km from the boma. Grazing herds were always monitored by one to several
herders and all livestock were kept within their enclosures when not grazing.
Illegal livestock grazing occurred within the Reserve, and sometimes occurred at

night, when Reserve rangers rarely patrolled.

Collection of conflict data

In February 2003, we trained three Maasai scouts, one from each of the
study locales, to complete a one-page report when notified of any injury or death
of livestock deemed to have been caused by carnivores. During the same
month, we held meetings with elders from each locale to discuss our project
goals and request their cooperation and assistance. We asked landowners to
inform their local scout of depredation events occurring either at the boma or
during grazing, as soon as possible after they occurred. Scouts visited each
attacked boma or herd owner to collect data. It was widely known by local
villagers that our research was not affiliated with the government or other political
entities, and that we offered no compensation for depredated livestock. There

was thus no apparent incentive for exaggerating or fabricating claims, yet scouts

137

made every effort to confirm all incidents based on available evidence. Bomas
more than 2 km from the Reserve were excluded to maximize efficient monitoring
of the area by scouts. A few settlements that contained no livestock of any type
were also excluded from the study. Conflict reports were collected from March
2003 through April 2004.

Based on available evidence and witness accounts, scouts recorded the
estimated time of day of each attack, the number and species of all livestock
killed or injured in the attack, the predator species involved, and the nature of the
interaction, if any, between villagers and the predator. In addition, a narrative
account of each event was recorded, as well as the evidence used to identify the
predator species. We documented the number of cattle held in the central
enclosure as well as the number of sheep and goats held in the owner’s small
stock enclosure. With the exception of the enclosures themselves, village
residents and their domestic dogs are the only deterrents to predator attack at
Maasai bomas. Neither firearms nor night watchmen are used in this region. We
therefore recorded the number of dogs associated with each affected household
as well as the total number of dogs at the boma. As an indicator of the number
of residents and the corresponding levels of human activity in attacked bomas,
the number of household huts and the total number of huts in the boma were
recorded. Finally, we categorized the type of enclosure used to protect small
stock and cattle as pole fence, bush fence, or other. All of the above variables
were recorded within 1—3 days of the incident. These same variables were also

recorded for all bomas and enclosures that did not suffer any depredation losses

138

during the study. A single observer (JMK) assessed the strength (strong vs.
weak) of attacked and unattacked enclosure fences based on their relative levels
of maintenance, reinforcement, and overall sturdiness.

The spatial locations of all study bomas and enclosures were recorded
using a hand-held GPS unit. We measured the distance from each boma and
enclosure to the nearest vegetative cover adequate to conceal a predator in
daylight, because predators may be less willing to attack livestock further from
cover. Because densities of some large predators are likely lower outside than
inside protected areas (Caro 1999b; Mills 8 Hofer 1998; Ogutu et al. 2005),
bomas further from the Reserve may be less vulnerable to attack. We therefore
measured the distance to the Reserve border for each study boma and enclosure
using ArcView GIS 3.2 (Environmental Systems Research Institute, Redlands,
CA, USA). Finally, the isolation of bomas relative to other bomas may also
influence their vulnerability to attack. We therefore recorded the distance to the
nearest boma, and the density of bomas within a 200 m radius of each boma.
The same analysis was repeated using enclosure locations to characterize the
relative isolation of individual sheep/goat enclosures. Accurate locational data

were unavailable for attacks on grazing herds.

Ecological conditions
Previous research has suggested natural prey abundance may influence
depredation rates (Polisar et al. 2003; Woodroffe et al. 2005), and that rainfall

may be an indirect measure of prey abundance and observed variation in

139

depredation frequency (Patterson et al. 2004; Woodroffe 8 Frank 2005). We
therefore examined relationships between temporal variation in depredation
frequency and both rainfall and prey abundance. Total monthly rainfall was
measured using a standard metric rain-gauge located along the Talek River. We
assessed the availability of natural prey to large predators by counting all prey
occurring along 29 1-km road transects, two of which were located on group
ranch property (Figure 4.1). We counted all wild ungulates within 100 m of each
transect 2—4 times per month for 13 of 14 months during the study period. An
average number of total ungulates counted per census was then calculated as an

index of local prey abundance in each month.

Predator movements

We monitored radiocollared adult spotted hyenas throughout the study
period to determine whether predator movements were associated with temporal
variation in depredation behavior. Spotted hyenas live in social groups called
clans and cooperatively defend a stable group territory. Monitored hyenas were
members of a single clan whose northern territory boundary extended into the
Talek and Ntipilikwani locales outside the Reserve, and whose 47—55 members
were known to be involved in local depredation events. Between 2001 and 2005,
at least nine individuals from this clan were killed at bomas within the study
region during livestock attacks.

We documented hyena space use with two different monitoring

techniques. The first method utilized frequent (~ 1 location per hyena every 2—3

140

days) telemetry locations collected at all times of day and night, with the majority
of monitoring effort occurring near dusk and dawn. Three individual home-
ranges (HRs) were constructed for each of 8 hyenas (4F, 4M) based on a
minimum of 35 locations per hyena ()7 = 57 locations) collected during months in
each of 3 depredation categories. Months having < 4 hyena attacks were
classified as “low” depredation periods, between 4 and 7 as “mid”, and > 7 as
“high”. We then associated each depredation category with the proportion of
each individual’s corresponding HR situated outside the Reserve. All home-
ranges were calculated with Animal Movement Analyst (Hooge 8 Eichenlaub
2000) as 95% fixed-kernel utilization contours with smoothing factors (h)
determined using least-squares cross-validation (Seaman 8 Powell 1996; Worton
1989)

Our second method utilized long-term (2—15 hr) follows of nine
radiocollared hyenas (6F, 3M) conducted at all times of day and night. During
follows, locations of the focal hyena were collected every 10 min using telemetry,
often with visual confirmation, to assess the frequency of use by hyenas of the
group ranch properties outside the Reserve. The average proportion of locations
per follow on ranch property was compared to the proportion of the clan territory
extending into the ranches. The clan territory boundary was based on a 95%
fixed-kernel utilization contour constructed using 4763 locations of 11 adult
female hyenas collected from May 2002-April 2004. Infrared spotlights and
night-vision goggles were used to observe hyenas at night with minimal

disturbance. Special attention was paid to the behavior of followed hyenas in

141

close proximity to bomas or humans. For both follow-based and tracking-based
analyses, all locations collected within 200m of the communal den were excluded
from consideration to control for the influence of this location. Cubs less than
approximately 9 months of age, from all mothers in the clan, reside at the

communal den and it serves as a center of social activity for clan members.

Statistical analysis

Much of the data involving livestock losses and characteristics of attacks
were summarized using proportions, which were compared between two groups
using Fisher exact tests (Zar 1999). Relationships between monthly mean
values for ecological variables and monthly attack frequencies were first
investigated using Pearson’s correlation coefficients (rp). Data on monthly prey
abundance were log-transfonned to obtain normality. We then modeled both
rainfall and prey abundance in a multiple regression analysis, with monthly attack
frequency as the dependent variable. The relative influence of these two
variables on attack frequency was examined using partial correlation coefficients
(Bart et al. 1998). To identify whether hyena movements were associated with
specific ecological conditions, we used these same independent variables in a
regression model with % of HR outside the Reserve as the dependent variable.
For this model, locations from all monitored hyenas were pooled by month (3? =
120 locations per month, minimum = 74) to obtain a continuous monthly measure
of clan space use. To associate hyena depredation behavior with hyena space

use, average % HR outside the Reserve for individual hyenas was compared

142

among the three hyena depredation categories using the non-parametric
Friedman test for repeated measures.

We used univariate analyses to compare characteristics of bomas and
enclosures containing livestock that were attacked by predators with those of
bomas and enclosures that were not attacked. Most data describing boma and
enclosure characteristics were not normally distributed, so we used Mann—
Whitney U-tests to compare continuous variables between groups.

We next used these descriptive variables in multivariate logistic regression
analyses to determine which were useful in predicting probabilities of hyena and
leopard attacks on both bomas and individual sheep/goat enclosures for a total of
four model-building progressions. The set of predictor variables was initially
reduced by eliminating highly correlated continuous variables from consideration
using Spearman rank order correlation coefficients (rs). One of the variables in
each correlated (rs > 0.70) pair was excluded based on the results of exploratory .
univariate tests. All possible logistic regression models for each of the four
dependent variables, utilizing all combinations of the remaining predictor
variables, were then analyzed and compared using Akaike’s Information Criterion
(AlCc) values corrected for low sample sizes. Relative to the model with the
lowest AlCc value, models with a difference in AICc > 2.0 are considered to have
substantially lower empirical support (Burnham 8 Anderson 2002). We therefore
considered all models within 2.0 AlCc points of the lowest AlCc model. The
significance of individual logistic regression models was assessed using a

likelihood-ratio X2 test, while significance of model parameters was assessed

143

using Wald’s X2 test. All tests were considered statistically significant at on = 0.05,
and all statistical analyses were conducted using the software package

STATISTICA (StatSoft 2002).

RESULTS
Losses of livestock and carnivores

A total of 130 depredation events were recorded from March 2003 through
April 2004. Of the 104 monitored livestock enclosures, 48% suffered loss or
injury of stock due to predators during our study period. Every incident was
attributed to a specific predator, with 71% based on visual confirmation of the
predator, and the rest based on tracks, claw marks, or the condition of the
livestock carcass. Hyenas were involved in 69 of the 130 reported incidents
(53%), with leopards and lions involved in 32% and 15%, respectively. There
were no reported depredation events involving other predators.

During attacks, carnivores killed 147 stock animals; 115 sheep and goats
(78%), 30 cattle (20%) and two donkeys. Hyenas, leopards and lions were
responsible for 50% (n = 74), 37% (n = 55) and 12% (n = 18) of the livestock
deaths, respectively (Figure 4.3). Non-fatal attacks injured an additional 32
sheep and goats, 12 cattle, and one donkey. Leopards accounted for 48% of all
sheep and goats killed, but never attacked cattle (Figure 4.3). Lions killed only
one goat, but accounted for 57% (n = 17) of all cattle kills. Sheep and goats
comprised 80% of all livestock kills by hyenas. Hyenas were responsible for 51%

and 43% of all sheep/goat and cattle depredation respectively (Figure 4.3).

144

70 — - All Livestock

- Sheep and Goats
» Cattle
60 .
(D
C
.2
.o—a
(U
"U
(D
L—
Q.
(D
D
E
H
O
l.
M—
O
89

 

 

Hyena Lion Leopard

Figure 4.3. The percent of the total losses of sheep and goats, cattle, and all
livestock species that were attributed to spotted hyena, lion, and leopard
depredation. Percentages shown are out of a total of 147 livestock killed during a
14-month period along the northeast border of the Masai Mara National Reserve.

Kenya.

145

Seventy-five (58%) of the 130 recorded attacks occurred inside bomas,
with the remaining attacks directed at grazing animals. Hyenas attacked grazing
herds as often as corralled herds (45% and 55% respectively; Fisher exact test p
= 0.336) and were responsible for more than 80% of attacks both on cattle within
the boma, and on grazing herds of sheep and goats (Figure 4.4). Lions attacked
corralled livestock (n = 2) less than they attacked grazing herds (n = 18; Fisher
exact test p = 0.014), and were involved in 74% of attacks on grazing cattle herds
(Figure 4.4). Conversely, leopards attacked grazing herds (n = 5) less than
livestock in bomas (n = 36; Fisher exact test, p = 0.0003) and were responsible
for 56% of all attacks on sheep and goats in bomas (Figure 4.4). All attacks at
the boma took place during the night whereas 71% of attacks on grazing herds
took place between 11:00 and 16:00 h.

Of 109 fatal attacks, 81% resulted in the death of only one stock animal,
and 8% in the death of 3 or more. Although leopards tended to kill more
livestock per attack (1.34) than hyenas (1.06) and lions (0.9), the difference was
not significant (Kruskal-Wallis test H1130 = 2.27, p = 0.322). Most (76%) of the
attacks on livestock in bomas were detected in progress, often resulting in the
predator being chased from the scene. Undetected boma attacks were more
likely than detected attacks to result in the death of livestock (100% vs. 79%,
Fisher exact test, one-tailed p = 0.027). Although one of the assumed benefits of
the presence of dogs is their ability to alert residents to the presence of
predators, locations at which predators were detected during attacks were not

characterized by more dogs either at the attacked enclosure (U = 505.0, p =

146

100 - - Lion
- Leopard
l Hyena

80-

% of Attacks

 

 

Cattle-B (12) Cattle-G (23)
Sheep/goats—B (63) Sheep/goats-G (31)

Figure 4.4. Involvement of lion, leopard and hyena in 4 categories of livestock
depredation incidents along the northeast border of the Masai Mara National
Reserve during a 14-month study period. Total numbers of attacks of each type
are listed in parentheses after category headings. "B" indicates attacks at the

boma; "G" indicates attacks while grazing.

147

0.921) or at the boma (U = 457.5, p = 0.984) than were locations where attacking
predators went undetected. In addition, there was no difference in the number of
houses within either the homestead (U = 415.0, p = 0.982) or the boma (U =
418.5, p = 0.380) between locations of detected and undetected attacks. Only
one reported attack during the study period resulted in the death of the intruding
carnivore (a hyena was killed). In other cases, three hyenas and one lion were
speared during attacks, but escaped to unknown fates.

The number of sheep and goats reported for 99 individual enclosures was
estimated at 16,523 animals. Thus, the 115 sheep and goats killed by predators
resulted in an annual loss of 98.4 individuals, or 0.6%, of the study region’s small
stock holdings. Considering only the 48 owners that suffered sheep or goat
losses to predators, each suffered an average annual loss of 1.8% of his stock
(range: 0.2—8.6%). Given an estimated 11,864 cattle at the 78 study bomas, we
recorded an annual loss to predators of 25.7 cattle, or 0.2% of the total cattle
holding.

Based on local information from numerous sources, adult cows were
priced at 10,000 Kenya shillings (KSh), adult goats and sheep at 2,000 KSh,
calves at 6,000 KSh, and juvenile goats and sheep at 1,000 KSh. An exchange
rate of 76.04 to USD was used, based on the average exchange rate during the
study period (www.centralbank.g9.ke - accessed 6 Apr 2005). Over a 14-month
period, livestock depredation by predators resulted in a loss of 6,049 USD
(460,000 KSh) to the study region. Hyenas were responsible for 45% of this

monetary loss, lions for 36%, and leopards for 19%.

148

Temporal patterns of attacks and relationship to predator movements

There was substantial monthly variation in the number of depredation
events, with attack frequency highest from March — May and lowest in October
(Figure 4.5). Monthly attack frequency was positively correlated with total
monthly rainfall (rp = 0.66, p = 0.010, n = 14; Figure 4.6), and the two months
during which depredations reached their highest levels were the only months in
the study period during which rainfall exceeded 200mm. In addition, monthly
prey levels were weakly correlated with attack frequencies (rp = -0.47 p = 0.103;
n = 13 months): However, despite occurring during the influx of migratory
wildebeest and zebra from the Serengeti, October 2003 prey counts failed to
reflect this super-abundance of prey and indicated below-average prey numbers.
With this outlier month excluded, prey abundance showed a strong negative
correlation with attack frequency (rp = -0.67, p = 0.018; n = 12). Although both
variables were related to attack frequency, total monthly rainfall and average
monthly prey abundance were not correlated (r,D = -0.44, p = 0.149; outlier
excluded).

It was the belief of local livestock owners that not only was the rainy
season a time of increased predator conflict, but also that predators specifically
preferred to attack bomas during rainfall events. To investigate whether the
above correlations were driven by seasonal or nightly factors, we documented
the proportion of attacks that occurred on night when rainfall was recorded.

There was no difference between the proportion of nights during the study period

149

18

16

O)

# of Depredation Events
A

N

 

8I

0
MAMJJASONDJFMA

Month
Figure 4.5. Total monthly livestock attacks by predators recorded within 2 km of
the northeastern border of the Masai Mara National Reserve, Kenya, from March
2003 to April 2004. Conflicts included attacks by lions, leopards, and hyenas that

resulted in either death or injury of livestock.

150

_x _\ _x _L
N A O) m
.

Number of Conflicts per Month

0 1 l l l 1 n 1 L L l_ n _;
0 20 40 60 80 100 120 140 160 180 200 220 240

Monthly Rainfall (mm)

 

 

Figure 4.6. Correlation (rp = 0.66, p = 0.010) between monthly rainfall (mm) and
total predator incident reports collected each month during a 14-month study

penod.

151

on which it rained (30.1%), and the proportion of attack nights on which it rained
(39.4%, X2 = 2.692, p = 0.102). Thus, the correlation between monthly conflict
frequency and rainfall patterns does not appear to be driven by a predator
preference for raiding bomas on rainy nights.

Using multiple regression analysis, we compared the relative influence of
rainfall and prey abundance on attack frequency. The two-variable regression
model explained a significant amount of the variation in monthly attack frequency
(F(2,10) = 4.84, R2 = 0.492, p = 0.034), yet prey abundance explained only 9.1% of
the variance beyond that explained by rainfall. Thus, although both independent
variables were useful in predicting monthly attack rate, rainfall was a slightly
better predictor than local prey abundance.

Space use by radiocollared hyenas was related to hyena depredation
behavior. We found that the % of each hyena’s HR falling outside the Reserve
border was significantly higher in mid- (n = 5 months) and high-level (n = 4
months) depredation periods than in low-level (n = 5 months) periods (Friedman
X2 = 12.0, p < 0.003; Figure 4.7). This variation was not likely to have resulted
from changes in natural prey abundance outside the Reserve, as group ranch
prey transects showed negligible prey numbers throughout the year. On
average, 2.4 i 3.4 (SD) prey were counted per transect on group ranches each
month, compared to 40.6 i 31.2 prey per transect inside the Reserve. Neither
rainfall nor prey abundance was useful in predicting the % of monthly clan HR

outside the Reserve (5210, = 1.19, R2 = 0.192, p = 0.344).

152

_L
O)

 

_L
A

_s
M

_.x
C)

 

 

 

[3

2f _L

Mean % of HR outside Reserve
CD
I
l

 

 

 

Low Mid High
Monthly Depredation Frequency

Figure 4.7. Mean percent of individual hyena home ranges (95% fixed-kernel)
that were located outside the Masai Mara National Reserve based on locations
collected during months of low, medium and high frequencies of livestock
depredation events involving hyenas from March 2003—April 2004. Whiskers

indicate 95% confidence intervals.

153

We collected 1754 locations (n = 811 night, 943 daylight) during multiple
follows on each of 9 different hyenas. Both nighttime (n = 33) and daytime
follows (n = 39) provided an average of 24 locations per follow. Followed hyenas
frequently used of lands outside the Reserve. Although 9% of the clan territory
lay outside the Reserve, an average of 17% of locations per nighttime follow
were collected outside the Reserve. However, on average, only 0.4% of
locations per daytime follow were outside the Reserve. Of 130 locations
recorded outside the Reserve during all follows, 23% were within 200 m of a
boma. When outside the Reserve, followed hyenas were often seen foraging in
close proximity to bomas, but making no attempts to enter them. In addition, our
observations suggest that, at least during hours of darkness, hyenas do not
appear to be concerned about humans. For example, groups of hyenas
sometimes slept for extended periods within 150 m of large bomas. In addition,
hyenas were seen to walk calmly within 50 meters of humans, only fleeing from

those carrying flashlights.

Characteristics of bomas and enclosures
Univariate Compan'sons

Bomas that suffered at least one hyena attack differed from those that did
not with respect to five of nine variables (Table 4.1). Bomas attacked by hyenas
contained larger numbers of cattle, sheep and goats, sheep/goat enclosures,
dogs and houses than did bomas not attacked by hyenas. All significant

variables with the exception of the number of cows at the boma were

154

88 u 8 8 88588 C

mmfimtg 86.85885 C

 

 

88.0 8.8 9 8.82 Nov 9. :5 8.0 8.8 9 4.88 88 9 3mm 8 c5 8&8 88.0
88.0 no 9 EN 3 9 so 83 no 9 3 $.83 88 E 08 s 88m
83 v8 9 38 «.8 h 885 83 Se 9 8.88 3.88.88 8 A8 .98 s .85
80o to 9 3 ed 9. 5 888 od 9 3 8.93 E .888 88888
88 :8 9 38 we: 9 32 888 Km, 9. 88 6.83. E 8 as 888m 9 .85
83 no 9. N8 3 9 mm 88.0 to 9 3 a. 8.8 8 .888:
88.0 to n we we 9 m8 838 to 9 we 8. :88 8 .88
88 8.8 H 8.88 8.88.88 858 8.28. E 8.88.88 8 888888
owed 298:.me 3.3882 88.0 E: :38 3.88.88 8 850
m:_m>-q oz 8; c:_m>-q oz wm> z

 

 

xomz< mEom 2803

xomz< mEom 29f

.nmEcEOBBE 8m £8.93 3 8:86.: mmEmtm> .992 D >mEE>>.ccm_2 co umwmn 2m mm:_m>d 8me 628851

_mcozmz 95.2 _mme 85 v6 .6an Emvmmmcto: 85 v6 Ex N 55318353 mmEon _mwmms. noxomzmca new Umxomum

am 68288 853:? EcncmacuE 9:: Co .0th Emncfim H mc:_m> cmmE $928 8086an0 9m_._m>_c3. :4 was.

155

intercorrelated (rs > 0.70; Table 4.1). Bomas suffering at least one leopard attack
on livestock had fewer other bomas within a 200 m radius, and were further from
the closest boma than those suffering no leopard attacks (Table 4.1).

Few differences were identified between attacked and unattacked
sheep/goat enclosures (Table 4.2). Enclosures attacked by hyenas were closer
to the next enclosure and were more frequently constructed of local bush
material than were unattacked enclosures (Table 4.2). Enclosures suffering
leopard attacks only differed from unattacked enclosures with respect to fence
type, with attacked enclosures more likely to be made from pole fencing than
were unattacked enclosures (Table 4.2).

There was no difference between the proportion of pole (43%) and bush
fences (36%) present in the study area that were attacked by predators (Fisher
exact test, p = 0.656). However, sheep and goats enclosed by pole fences were
more likely to be attacked by leopards than were those enclosed by bush fences
(Fisher exact test, p = 0.002), whereas small stock within bush fences were more
likely to be attacked by hyenas than were those in pole enclosures (Fisher exact
test, p = 0.003; Figure 4.8). Livestock held under strong, well-maintained pole
fences were no less likely to be attacked by leopards (Fisher exact test, p =
0.752) or hyenas (p = 0.286) than were those within weak pole fences. Similarly,
relative strength of bush fences did not affect probability of attack by either

hyenas (Fisher exact test, p = 0.757) or leopards (p = 1.0).

156

8.0 u e 8 88588 C

 

830.0 mém NNN
moood va mu?
wtd veNHNemw YEVHQFON
Nmod md H Nd md H NN
mood md H @274 0.0 H we

mmvd md + 5m md + 0.8

med EN? H Yon; Vém H QmON

«mood me MEN

Rood Now We

820d ©.m_. H 009. ohm H No:
oomd v.0 H QN md H Em
momd md H we v.0 H 08

mend m.o + Em md + 06

mwwd I: H 9me ”mm H ”mm?

mm 3% 858.05 swam
mm 3% 938.05 m_on_
mm AEV 938.08.. 6320
8 E 8 com 5 858.88
om 95 238.05 Cm .0686:

mm 95 938.88 E 800

8 E 885 888888

 

m:_m>-q oz 8;

 

x033. 6006885 2300..

m:_m>-q oz wm>

 

xomt< 5.80068me mag:

Z

6986:0825 macs mum; ucEmem mm_nm:m> 85 ,6 8:02 .mcoanoa

Lou. E8: 698 8.2“. can 838:? wzozczcoo .8 988 b >mSE>>écm§ :0 Emma 98 mc:_m>d .963. .chwmm

_mcozmz Ems. 58.2 05 yo .mEon Emwmmmctoc 85 3 Ex N £55, $58.05 809885 89.08th new @3033

E 82:89 8:98:85 Eocccamng co>cm n_o Etc Emncfim H mm:_m> cmmE cmmszmn mcowcmano 38:82:: .Né 2an

157

1.0 . - Pole sheep/goat fence (n=40)
- Bush sheep/goat fence (n=44)

0.8 -

% of Available Fences

 

 

Attacked by Hyena Not Attacked
Attacked by Leopard

Figure 4.8. Relative attack rates by leopard and hyena on the two most common
types of sheep/goat enclosures within 2 km of the northeastern Reserve border.
Significant differences (Fisher exact test for comparison of proportions; p < 0.05)

are indicated by an asterisk.

158

Multivariate Analyses

Nine variables were initially considered for estimation of hyena and
leopard attack probability at Maasai bomas (Table 4.1). The most significant of
all intercorrelated variables, number of boma houses, was retained, but numbers
of sheep and goats, sheep/goat enclosures, and boma dogs were excluded from
further analysis. With occurrence of a hyena attack as the dependent variable,
all possible model combinations of the six remaining variables were compared
based on model AlCc values (Table A20). The lowest AICc value was assigned
to a model including the number of boma houses (Wald’s 12 = 13.52; p < 0.001)
and the distance to the Reserve (Wald’s X2 = 3.92; p = 0.048) as predictive
variables (Log-likelihood X2 = 22.54, p < 0.0001; Table 4.3). Two additional
models were supported by the data, both of which combined these first two
variables with an additional third variable (distance to the closest boma or
number of bomas within 200 m). However, none of the variables other than
distance to the Reserve and the number of bomas houses were significant
parameters in the two additional models. Model selection procedures thus
indicated that likelihood of hyena attack on bomas increased as number of
houses increased and distance to the Reserve decreased.

Considering all possible models with the same six variables predicting
boma attack by leopards (Table A.21), the lowest AICc value was assigned to a
model including only the number of bomas within a 200 m radius (Wald’s x2 =
7.38, p = 0.007; Log-likelihood X2 = 12.17, p < 0.001; Table 4.3). Two additional

models were included in the optimal subset of models; both included the number

159

Table 4.3. Logistic regression models predicting the probability that a Maasai
boma (village) will be attacked by large carnivores. Presented models were
selected based on the minimum AIC value of all model combinations using 6

uncorrelated continuous variables.

 

Estimate St. Error Wald Stat. p-value
Blane
Intercept - 2.0664 0.8505 5.9027 0.0151
Boma houses (#) +0.3955 0.1075 13.5264 0.0002
Distance to Reserve (m) - 0.0016 0.0008 3.9226 0.0476
Legpflg
Intercept - 0.2393 0.3438 0.4847 0.4863
Bomas within 200 m (#) - 0.7476 0.2753 7.3760 0.0066

 

160

of bomas within 200 m, and added a single variable (distance to the Reserve or
the number of boma houses). However, none of these other models included
significant parameter estimates additional to the number of bomas within 200 m
(Table A.21). Therefore modeling procedures indicated that a decrease in the
density of surrounding bomas was the most important factor increasing the
probability of a leopard attack.

Finally, we attempted to identify variables most important in estimating the
probability of leopard and hyena attacks at the level of individual sheep/goat
enclosures. Both logistic regression models initially considered five continuous
variables and one categorical variable representing fence type. Model-building
for hyena enclosure attack probability indicated a set of eight optimal models
(Table A22). The lowest AlCc value was assigned to a model that included only
fence type (Wald’s X2 = 9.64; p = 0.002) as an independent variable (Log-
likelihood X2 = 6.42, p = 0.011; Table 4.4). No additional variables in the other
seven models were significant model parameters (Table A22). The odds ratio

for fence type indicated that the presence of a bush fence increased the
. probability of hyena attack by 2.43 times. With leopard enclosure attack as the
dependent variable, we identified three optimal models (Table A23). The model

with the lowest AICc only included fence type (Wald’s X2 = 11.88; p = 0.001) as a

model variable (Log-likelihood X2 = 11.49, p = 0.001; Table 4.4), and no
additional variables in the other two models were significant model parameters
(Table A23). The odds ratio for fence type indicated that the presence of a pole

fence increased the probability of leopard attack by 2.67 times.

161

Table 4.4. Logistic regression models predicting the probability that a Maasai
sheep/goat enclosure will be attacked by large carnivores. Presented models
were selected based on the minimum AIC value of all model combinations using

five uncorrelated continuous variables and one categorical variable.

 

Estimate St. Error Wald Stat. p-value
diam
Intercept +0.7403 0.5737 1.6653 0.1969
99828928? ______________ :‘R-E‘EYE _______ 93.82? _________ 9 ;§:‘P_§____-_-2-_9933___
m
Intercept +1.2820 0.2842 20.3456 0.0000
Pole fence? +0.9797 0.2842 11.8823 0.0006

 

162

DISCUSSION

Predator Involvement and Livestock Losses

Our study was designed, in part, to complement and expand on recent
work on livestock depredation by carnivores on rangelands in eastern Africa.
Patterson et al. (2004), who conducted their study on commercial ranches in
southeastern Kenya, found lions to be responsible for 86% of attacks on
livestock, with hyenas involved in <10%. They reported no leopard attacks on
livestock. A study conducted on primarily commercial ranches in northern
Kenya, found that lions accounted for approximately 63% of all livestock kills,
with hyenas and leopards accounting for only 15% and 11% of kills, respectively
(Ogada et al. 2003). In an earlier study conducted on group ranches near our
own study site, Karani (1994) found that leopards were the most serious livestock
predators (50% of livestock attacks), with lions and hyenas responsible for 31%
and 19% of recorded attacks respectively. Thus, although multiple earlier studies
on Kenyan rangelands concluded that the lion is the most serious livestock
predator, and that hyena predation is relatively infrequent (Frank 2000; Ogada et
al. 2003; Patterson et al. 2004), we found relatively little involvement by lions in
livestock attacks, particularly at Maasai bomas, with leopards and hyenas
responsible for most attacks.

Regional variation in relative livestock depredation by these large
predators may be attributed to differences in relative densities of large
carnivores, husbandry practices, and relative abundance of different stock

species. Although some researchers have been unable to associate predator

163

density with livestock depredation rates (Connor et al. 1998; Graham et al. 2005),
others have clearly documented increases in depredation with increases in
carnivore density (Sagor et al. 1997; Stahl et al. 2001; Stoddart et al. 2001).
Recent surveys on the Koyake group ranch have indicated that lion densities
there may be very low (Ogutu et al. 2005). However we assumed at least some
of our depredation events involved lions from inside the Reserve, where lion
density was relatively high (0.369 Iions/kmz; Ogutu et al. 2005). Hyena density in
the northeastern portion of the Reserve was estimated to be 0.86/km2 (Frank
1986a), one of the highest densities reported in Africa. Unfortunately predator
densities are not reported in most studies, preventing direct comparisons.
Husbandry practices on commercial ranches may reduce the relative
involvement of hyenas in livestock depredation. Various researchers have
concluded that rates of livestock loss to predators in Kenya, particularly hyenas,
could be reduced through construction of sturdier boma fences (Frank 2000;
Kruuk 1981), and bush fences for livestock corrals on commercial ranches are
often sturdier than those built in pastoral bomas (Ogada et al. 2003). However,
Ogata et al. (2003) found no effect of boma height or thickness on depredation
rates. Our data support the conclusion that improved fencing, at least on
pastoral ranches, is not necessarily an effective solution to livestock depredation.
Finally, relative availability of small and large stock animals is also likely to
influence involvement of predators in livestock depredation. The Patterson et al.
(2004) study, which reported low hyena depredation, included ranches on which

the majority of stock animals were cattle (Patterson et al. 2004). The low

164

frequency of hyena and leopard depredation in some areas may thus result from
the rarity of their preferred livestock prey, sheep and goats.

Hyenas attacked bomas approximately once every 11 nights. Although
densities of both lions and hyenas inside the Reserve appear to be high, lions
rarely attacked livestock in bomas outside the Reserve. Perhaps lions prefer to
remain in their defended territories within the Reserve even when prey
abundance there is relatively low, as even at these times natural prey abundance
inside the Reserve is far greater than on our study group ranches.

The annual loss of 0.6% and 0.2% of the total small stock and cattle
holdings respectively for our study region falls within the range reported for a
large subset of depredation studies from around the globe (0.02% to 2.6% of
local livestock holdings Graham et al. (2005)). Within Kenya, reported annual
livestock losses to predators range from 0.7% to 5.5% (Frank 1998; Karani 1994;
Kruuk 1981; Patterson et al. 2004), indicating that our observed depredation
rates are relatively low for Kenyan rangelands. In contrast to Ogada et al.
(2003), who found cheetah to be a significant predator of sheep and goats (49%
of 195 sheep and goat kills away from the boma), we found no cheetah predation
on livestock. This, together with the relatively small impact of lions on livestock,
particularly sheep and goats, may account for our low depredation rates.
However, these low annual stock losses to predators fail to represent the
significance and costs of depredation events to individual owners, who have
been known, in our study area, to lose up to 70 sheep and goats in a single

attack by hyenas. Such a loss to an individual livestock owner is catastrophic

165

and can also result in devastating retaliatory attacks. For example, in 1990, at
least 16 hyenas were killed in a single poisoning event, following a depredation
event in our study area (Holekamp 8 Smale 1992). Our study documented very
few carnivore deaths during livestock attacks; however, the loss of even single
carnivore individuals can have important economic impacts on areas benefiting
from wildlife tourism (Sillero-Zubiri 8 Laurenson 2001). And, although Kenyans
are legally permitted to kill carnivores in defense of their livestock, these events
are generally underreported due to fears of government fines or penalties and

likely resulted in an underestimation of carnivore losses during our study period.

Attacks on grazing herds

Attacks on grazing herds here were equally as common as were attacks at
bomas. While previous studies on commercial ranches have found
approximately 25% of livestock attacks to occur on grazing herds (Ogada et al.
2003; Patterson et al. 2004), pastoral ranches have documented up to 90% of
predator attacks on livestock to be directed at grazing herds (Kruuk 1981).
Attacks on grazing herds are probably more frequent on pastoral group ranches
than on commercial ranches due to differences in herdsman behavior. In the
case of commercial ranches, herders are paid for their work, may be fired for
inadequate herd attendance, and work in groups large enough to discourage
stock theft (Ogada et al. 2003). Larger groups of herders appear to be effective
at limiting attacks on grazing herds (lkanda 2005; Ogada et al. 2003). Herds on

pastoral ranches however, such as those included in our study, are often

166

accompanied only by small groups of young boys who vary considerably in their
level of attendance to the herd.

Notably all three livestock predators in this study are primarily nocturnal,
yet many of the attacks on grazing herds took place during the middle of the day.
This indicates that not only are all three predator species bold enough to attack
livestock herds accompanied by herdsmen, but also that they are willing to attack
during periods when visibility of herdsmen is high. The preponderance of attacks
during midday may represent predatory behavior that capitalizes on a reduction
in herdsmen attentiveness and a corresponding increase in the spatial dispersion
of the herd that are likely to occur near the heat of midday. Hyenas in particular
have been shown to recognize and capitalize on similar behavior changes by
their willingness to attack sleeping topi during midday (Rainy & Rainy 1989).
Accounts of attacks on grazing herds in our study often described predators
rapidly emerging from vegetative cover to attack. We therefore suggest that
herdsmen avoid densely vegetated areas, where possible, particularly during the
rains when attacks are most frequent and vegetation is thickest. Studies that
investigate the spatial location of attacks on grazing herds in combination with
herdsmen behavior, herdsmen numbers, as well as herd size and composition
for attacked and unattacked herds are required to determine the effectiveness of

various husbandry techniques in minimizing depredation losses.

167

Temporal patterns of attacks and relationship to predator movements
Monthly rates of carnivore attacks on livestock were related to both rainfall
and natural prey abundance. Although Maasai pastoralists in some areas
recognize rainy seasons as periods of increased carnivore conflict (Patterson et
al. 2004; Rudnai 1979), some previous researchers have been unable to
associate rainfall with depredation frequency (Rudnai 1979), while others have
found the highest rates of depredation in the dry season (Butler 2000; lkanda
2005). However, elevated rates of lion-human conflict have been associated with
the monsoon rains in India, and recent studies have documented increases in
livestock depredation during the rains in Africa (Patterson et al. 2004; Woodroffe
& Frank 2005). Our data further support the importance of this relationship. Both
Patterson et al. (2004) and Woodroffe and Frank (2005) suggest this trend may
be ultimately driven by seasonal variation in local availability of natural prey,
which is dictated by rainfall patterns. Whether the wet or dry season brings
increased livestock depredation is likely ultimately dictated by the regional
relationship between rainfall and natural prey. Whereas the dry season in some
areas is associated with increased natural prey and reduced livestock
depredations, the inverse has been shown in areas where prey levels peak in the
wet season (e.g. southern Serengeti; lkanda (2005)). Although our data failed to
directly relate prey abundance with rainfall, prey abundance clearly influenced
depredation rates. Many studies have documented relatively high rates of
carnivore predation on livestock in areas occupied by few natural prey (Meriggi &

Lovari 1996; Mishra 1997; Polisar et al. 2003; V03 2000; Woodroffe et al. 2005),

168

while others, including this study, have shown depredation rates to increase as
natural prey abundance decreases over time (Stoddart et al. 2001).

Temporal variation in depredation behavior by hyenas was associated
with changes in space utilization by monitored hyenas. As expected, hyenas
used group ranch property more during months when hyena attacks on livestock
were most frequent. However, the ecological conditions associated with this
spatial shift were less obvious. We anticipated hyena movements would shift
toward the group ranches when rainfall increased and natural prey abundance
decreased. However, no correlation was identified between either of these
variables and HR shifts. Tracking data suggested that hyenas spent more time
outside the Reserve than expected, based on the small proportion of the clan’s
territory lying outside the Reserve, and that hyenas outside the Reserve were

often near bomas.

Boma and enclosure vulnerability

Both univariate and multivariate analyses indicated that, contrary to
previous findings (Ogada et al. 2003), increased human activity was associated
with an increased probability of hyena attack. Based on these and follow data,
we suspect that hyenas, as opportunistic feeders, are making regular visits to
bomas not for livestock primarily, but rather for discarded food and other edible
items. Large bomas, with more enclosures and houses, would thus be most
attractive to hyenas interested in exploiting refuse and opportunistic attacks on

livestock should therefore be more likely to occur at these bomas. Given the

169

attractiveness of these sites to foraging hyenas, secure refuse disposal at bomas
may reduce hyena attack frequency.

Unlike hyenas, leopards preferred to attack bomas that were relatively
isolated on the landscape. In contrast to hyenas, leopards generally only
consume fresh meat and are not known to frequent open spaces, such as those
surrounding most bomas in our study area, as they rely heavily on stealth and
ambush while hunting. Therefore, a leopard approaching a boma is much more
likely than a hyena to be searching specifically for livestock prey. While our
findings suggest that leopards avoid dense aggregations of human settlements,
they do not indicate leopards specifically select smaller bomas, as did the results
of Ogada et al. (2003). Our results may therefore suggest a trade-off in boma
selection by leopards. While isolated bomas offer a reduced level of human
activity and likely reduced probability of predator detection, bomas with fewer
enclosures or livestock offer reduced opportunities to access appropriately
vulnerable prey. As demonstrated by Ogada et al. (2003), dogs in our study area
did not seem effective in deterring leopard or hyena attacks. This is further
supported by reports from villagers suggesting that dogs were killed and eaten by
both predators with some frequency.

Surprisingly we found no effect of distance to cover on probability of attack
at a boma by either hyena or leopard. Ogada et al (2003) also found no
influence of cover on leopard and lion depredation rates at bomas, but did see a
trend for hyenas to attack bomas closer to cover more frequently. In general,

very few of the bomas in our study region were close to dense vegetative cover

170

with distances to cover averaging more than 250m. Predators in this region then,
may simply be used to crossing open space at night to approach villages.
Indeed, hyenas were observed to casually and slowly travel across large open
plains between villages without apparent distress.

The only variable effective in estimating the vulnerability of sheep/goat
enclosures was fence type. Overall, fences made from bush and pole material
were equally susceptible to predator attack. Given that material used to make
pole fences is expensive and not obtained locally, this was a surprising finding.
Pole fences offer certain advantages over bush fences in that they require less
maintenance and appear to be effective deterrents against hyena attack.
However, the use of pole material in enclosure construction more than doubled
the likelihood of a leopard attack. Although many villagers reinforced the pole
fences with iron sheeting, barbed wire or thorn bush to close gaps and remove
possible footholds, leopards appeared capable of capitalizing on small
weaknesses in these reinforcements. Bush fences however, seemed effective at
limiting leopard attacks, probably because they are often built at an outward
angle and provide few sturdy footholds for climbing. Unfortunately, the use of
bush material in enclosure construction more than doubled the likelihood of a
hyena attack, as hyenas proved highly adept at pushing through even the
densest of bush fences.

Given the absence of a relationship between predator attack frequency
and fence quality, and the time, labor, and depletion of local vegetation involved

in maintaining a strong fence (Kruuk 1981), it may be more effective for livestock

171

owners to concentrate efforts on developing novel methods of detection and
interruption of carnivore attacks, than on improving fences to minimize losses.
Our data further indicate that increased human activity not specifically designed
to deter predators, may be ineffective in reducing the probability of attack.
However, actively guarding bomas (e.g. posting night guards, sleeping in huts
within enclosures), particularly with the help of lights, may prove effective.
Investment of major effort in guarding enclosures, a practice rarely utilized in our
study area, may be most worthwhile during the rainy season, when attacks are
most common, and could potentially be relaxed when migratory herds are
present. Selection of fence type, however, is clearly important. Pole enclosures
are effective at minimizing losses to hyenas and should therefore be used, when
affordable, at larger bomas, particularly those built in close proximity to other
bomas, which our data indicate are more susceptible to hyena attack. Bush
fences, though still permeable to persistent leopards, seem to provide superior
leopard exclusion and should therefore be favored at isolated bomas, which
appear more vulnerable to leopard attacks. With respect to our finding that boma
isolation and size can play an important role in vulnerability to predator attack,
similar findings in North America regarding wolf depredation on cattle farms
(Mech et al. 2000) indicate that these factors may be important spatial predictors
of attacks on livestock not only by African predators, but by predators worldwide.
Our study has demonstrated that monitoring of both socio-ecological and
environmental variables, together with collection of detailed depredation

information, can be useful in generating practical recommendations for conflict

172

mitigation. In addition, knowledge of the movements and behavior of predators
involved in livestock depredation can offer important insight into the effectiveness
of prevention measures as well as the factors affecting temporal variation in

depredation rates.

173

APPENDIX

Table A1. Logistic regression results showing all models within 2.0 AIC points of

the lowest AlC model for all Mara River modeling periods based on hyena

locations collected during all times of day and night. Final selected models are in

bold. Only one model was identified for MR2 and MR3.

 

Model Variable Estimate Wald Stat. P AIC
MR1-Model #1 Dist to Den - 0.0009 243.10 <0.0001 2161.48
Dist to Stream - 0.0032 87.89 <0.0001
Prey Value +0.0508 30.06 <0.0001
Dist to Perm +0.0005 10.66 0.0011
Habitat 9.96 0.0069
Shrubland +0.3688 7.97 0.0048
Ripan'an F. - 0.4533 8.57 0.0034
MRiZMBHéi #‘2’ ' 'b'i's't't'é 'Dé’ri """ 3 "6.6666 """" 2'37 ."5' 8"" ' "'<'6.'6 6'0'1’ "'2’1'6'2121‘2'
Dist to Stream - 0.0033 80.95 <0.0001
Prey Value +0.0481 25.28 <0.0001
Habitat 10.01 0.0067
Shrubland +0.3809 8.41 0.0037
Riparian F. — 0.4426 8.10 0.0044
Dist to Perm +0.0005 8.29 0.0040
_____---_______--_D_i§_t_t.q.C.>9_v_er_-__:rD_.D_QQ;3_ __________ 1.08 ......... 0. 29.9.7. ___________
MR2 Dist to Den - 0.0007 190.56 <0.0001 -
Dist to Cover - 0.0017 38.03 <0.0001
Prey Value +0.0245 22.71 <0.0001
Dist to Stream - 0.0015 19.88 <0.0001
Habitat 15.00 0.0006
Shrubland +0.2337 5.50 0.0190
Riparian F. +0.1179 1.35 0.2459
-___-________-___-p_i§.t_t_q_F39.rm _____ til-9.99:4. __________ 9 .12 _________ 9 99.1.8. ___________
MR3 Dist to Den - 0.0009 596.97 <0.0001 —
Dist to Perm +0.0010 139.23 <0.0001
Dist to Cover - 0.0016 51.79 <0.0001
Dist to Stream +0.0017 44.78 <0.0001
Prey Value +0.0259 24.91 <0.0001
Habitat 19.97 0.0005
Shrubland - 0.0480 0.34 0.5582
Riparian F. +0.3147 14.93 0.0001

 

174

Table A2. Logistic regression results showing all models within 2.0 AIC points of

the lowest AIC model for the first Talek West modeling period (TKW1) based on

hyena locations collected during all times of day and night. The final selected

model is in bold.

 

Model Variable Estimate Wald Stat. P AIC
TKW1-Model #1 Dist to Den - 0.0014 201.27 <0.0001 3239.48
Habitat 123.15 <0.0001
Shrubland +0.6059 36.35 <0.0001
Riparian F. - 0.0439 0.07 0.7968
Dist to Perm +0.0006 22.94 <0.0001
Prey Value - 0.0430 22.17 <0.0001
____________________ Luv0016292200024
TKW1-Model #2 Dist to Den - 0.0014 198.34 <0.0001 3239.94
Habitat 96.97 <0.0001
Shrub/and +0.6381 37.70 <0.0001
Ripan'an F. - 0.0201 0.01 0.9067
Prey Value - 0.0424 21.59 <0.0001
Dist to Perm +0.0006 21.16 <0.0001
LUV - 0.0149 7.59 0.0059
____________________ 951299.292“__"429.9924---"Mt-.529---“-.9-_2_1.2_2_______--___
TKW1-Model #3 Dist to Den - 0.0014 176.83 <0.0001 3240.75
Habitat 97.69 <0.0001
Shrubland +0.6587 38.69 <0.0001
Ripan'an F. - 0.0583 0.11 0.7402
Dist to Perm +0.0006 22.14 <0.0001
Prey Value - 0.0428 21.97 <0.0001
LUV - 0.0164 8.58 0.0034
Dist to Cover +0.0004 2.19 0.1391
Dist to Stream - 0.0002 1.19 0.2762
'T'RWf-Tlin'éa‘e’l’fliimb'iéi’tb' béh""""‘-'6T6i)’f& """ T7234"""<’6.'66'o'1'"§2'&b'.'9‘2'
Habitat 119.21 <0.0001
Shrubland +0.6149 36.77 <0.0001
Riparian F. - 0.0727 0.17 0.6782
Dist to Perm +0.0006 22.82 <0.0001
Prey Value - 0.0434 22.44 <0.0001
LUV - 0.0174 9.72 0.0018
Dist to Stream - 0.0001 0.56 0.4535

 

175

Table A3. Logistic regression results showing all models within 2.0 AIC points of

the lowest AIC model for the second Talek West modeling period (TKW2) based

on hyena locations collected during all times of day and night. The final selected

model is in bold.

 

Model Variable Estimate Wald Stat. P AIC

TKW2-Model #1 Prey Value +0.0733 153.21 <0.0001 2009.42

Habitat 152.46 <0.0001

Shrubland +0.8041 39.53 <0.0001

Riparian F. +0.0489 0.05 0.8300

Dist to Den - 0.0009 57.76 <0.0001

Dist to Stream - 0.0023 40.11 <0.0001
____________________ PJ§I_t9_E9IEU___-__.t9;99_9§__-____lQ-21______.Q-.99_1.Q-----__---_
TKW2-Model #2 Prey Value +0.0737 143.61 <0.0001 2011.37

Habitat 110.65 <0.0001

Shrubland +0.8105 38.15 <0.0001

Riparian F. +0.0553 0.06 0.8099

Dist to Den - 0.0009 57.11 <0.0001

Dist to Stream - 0.0023 36.11 <0.0001

Dist to Perm +0.0005 10.95 0.0009
_______________ Pj§s.ts>_.<29_v.<-=:r______199922--______9_-9_§__----.9-_§z1€i_--____---_
TRWZ-Model #3 Habitat 151.62 <0.0001 2011.41

Shrubland +0.8049 39.22 <0.0001

Riparian F. +0.0480 0.04 0.8336

Prey Value +0.0734 151.27 <0.0001

Dist to Den - 0.0009 55.81 <0.0001

Dist to Stream - 0.0023 40.03 <0.0001

Dist to Perm +0.0005 10.62 0.0011

LUV - 0.0002 0.00 0.9515

 

176

Table A4 Logistic regression results showing all models within 2.0 NO points of

the lowest AIC model for the third Talek West modeling period (TKW3) based on

hyena locations collected during all times of day and night. The final selected

model is in bold.

 

Model Variable Estimate Wald Stat. P AIC
TKW3-Model #1 Dist to Den - 0.0007 40.35 <0.0001 1591.85
Habitat 34.73 <0.0001
Shrubland +0.7352 14.42 0.0001
Riparian F. - 0.4564 1.66 0.1976
Dist to Cover - 0.0025 17.48 <0.0001
Dist to Perm - 0.0004 14.02 0.0002
Dist to Stream +0.0010 9.67 0.0019
Prey Value +0.0214 2.10 0.1471
TKWSIM'da'é’I'fl'2"‘D'i§i'tb' ijé'n' """" - '070'0'08"""‘5’0.’1’0"""<'6.'000'1’ "159139"
Habitat 35.63 <0.0001
Shrubland +0.7688 15.69 <0.0001
Riparian F. - 0.5149 2.11 0.1462
Dist to Perm - 0.0005 24.06 <0.0001
Dist to Cover - 0.0026 19.15 <0.0001
Dist to Stream +0.0010 9.13 0.0025
____________________ £501."-----__-___-_-.9_-09_72_ _ .1_-§§-__---_9-_1_z_2§___-__-----
TKW3-Model #3 Dist to Den - 0.0007 """ 48.79 <0.0001 1591.89
Habitat 34.75 <0.0001
Shrubland +0.7455 14.84 0.0001
Riparian F. - 0.4801 1.84 0.1748
Dist to Perm - 0.0005 21.94 <0.0001
Dist to Cover - 0.0025 18.58 <0.0001
Dist to Stream +0.0011 10.62 0.0011
'T’kWé-‘M'éa'él'fif ’D‘iéi'tb' Bé'n' """"" - '0T00'0i """" 21'1'.'§é"""<'0 T0001" "T5152". 50'
Habitat 35.43 <0.0001
Shrubland +0.7564 15.14 0.0001
Riparian F. - 0.4896 1.90 0.1680
Dist to Cover - 0.0025 17.99 <0.0001
Dist to Perm - 0.0005 15.50 <0.0001
Dist to Stream +0.0010 8.56 0.0034
Prey Value +0.0180 1.42 0.2339
LUV - 0.0061 1.27 0.2601

 

177

Table A5. Logistic regression results showing all models within 2.0 AIC points of

the lowest AIC model for the fourth Talek West modeling period (TKW4) based

on hyena locations collected during all times of day and night. The final selected

model is in bold.

 

Model Variable Estimate Wald Stat. P AIC
TKW4-Model #1 Habitat 113.72 <0.0001 2729.17
Shrubland +1.1725 73.13 <0.0001
Riparian F. - 0.9631 16.12 <0.0001
Dist to Den - 0.0004 91.82 <0.0001
Dist to Perm - 0.0007 66.92 <0.0001
Dist to Cover +0.0020 34.78 <0.0001
Dist to Stream - 0.0005 3.18 0.0746
’T’kWiIIiriéa’éi #‘é'TiaBit'at """""""""""""" i’1'6’f1'5"""<'0.'000'1'"i'7'§0'.4§
Shrubland +1 . 1694 72.78 <0.0001
Riparian F. - 0.9341 15.22 <0.0001
Dist to Den - 0.0004 88.47 <0.0001
Dist to Perm - 0.0008 77.15 <0.0001
____________________ 91§3.t9_99.v.ar______3:095:19-__-_-_9.2.10--____<.9-_99_0.1.------___--
TKW4-Model #3 Habitat 112.73 <0.0001 2730.63
Shrubland +1.1673 72.33 <0.0001
Riparian F. - 0.9562 15.87 <0.0001
Dist to Den - 0.0004 87.05 <0.0001
Dist to Perm - 0.0007 59.97 <0.0001
Dist to Cover +0.0020 35.27 <0.0001
Dist to Stream - 0.0005 3.29 0.0695
____________________ EEK---_______--__.t9;90_%9 - _0-.55 - 0427.5
TKW4-Model #4 Habitat """" i132?” "<0.0001" 2751.14
Shrubland +1.1723 73.09 <0.0001
Riparian F. - 0.9593 15.87 <0.0001
Dist to Perm - 0.0007 64.88 <0.0001
Dist to Den - 0.0004 34.66 <0.0001
Dist to Cover +0.0020 31.27 <0.0001
Dist to Stream - 0.0005 3.08 0.0794
Prey Value +0.0009 0.03 0.8530

 

178

Table A6. Logistic regression results showing all models within 2.0 AlC points of

the lowest AIC model for the first Mara River modeling period (MR1) based on

hyena locations collected during grazing hours (0900-1800 h). The final selected

model is in bold.

 

Model Variable Estimate Wald Stat. P AIC
MR1-Model #1 Dist to Stream - 0.0070 72.18 <0.0001 725.18
Dist to Den - 0.0011 67.65 <0.0001
Dist to Perm +0.0011 13.08 0.0003
Habitat 11.69 0.0029
Shrubland +0.7354 1 1.20 0.0008
Riparian F. - 0.5324 2.91 0.0878
Dist to Cover +0.0009 3.44 0.0635
____-____---____-_P_[e.¥.\_/?_'9§ ______ r6666? ___________ 2. .1? _________ Q .1495 ___________
MR1-Model #2 Dist to Den - 0.0012 81.98 <0.0001 725.33
Dist to Stream - 0.0070 71.66 <0.0001
Dist to Perm +0.0012 17.40 <0.0001
Habitat 11.74 0.0028
Shrubland +0.7316 10.99 0.0009
Riparian F. - 0.4866 2.47 0.1163
_-_-_--__--_------61631666265--16-9.61.1 ___________ 6 .2? _________ 6 61.2.6 ___________
MR1-Model #3 Dist to Den - 0.0010 73.53 <0.0001 726.44
Dist to Stream - 0.0064 71.85 <0.0001
Dist to Perm +0.0011 15.74 <0.0001
Habitat 10.55 0.0051
Shrubland +0.7001 10.36 0.0013
Riparian F. - 0.5886 3.64 0.0564
_-_-_--__-___-____P.r.6x.v6.ly_6 ______ tit-9.4.66. __________ 4. 16 _________ 6 62.9.6 ___________
MR1-Model #4 Dist to Den - 0.0011 86.48 <0.0001 728.99
Dist to Stream - 0.0061 67.67 <0.0001
Dist to Perm +0.0013 23.66 <0.0001
Habitat 10.03 0.0066
Shrubland +0.6803 9.74 0.0018
Riparian F. - 0.5435 3.15 0.0760

 

179

Table A]. Logistic regression results showing all models within 2.0 AIC points of

the lowest AIC model for the first Mara River modeling period (MR1) based on

hyena locations collected during livestock-free hours (1900—0800 h). The final

selected model is in bold.

 

Model Variable Estimate Wald Stat. P AIC
MR1-Model #1 Dist to Den - 0.0011 159.48 <0.0001 1338.23
Prey Value +0.0521 19.39 <0.0001
Dist to Stream - 0.0015 14.86 0.0001
MEIR/.3212." 42’ "64113 3;; """ 363631”"EB?"""Zéfébb’imiééé'f7'7'
Dist to Stream - 0.0018 14.37 0.0002
Prey Value +0.0470 14.24 0.0002
Dist to Cover +0.0004 1.49 0.2224
MR1-Model #3 Dist to Den - 0.0011 147.58 <0.0001 1339.06
Prey Value +0.0485 15.65 <0.0001
Dist to Stream - 0.0017 14.62 0.0001
__-_-_----___-___-P.6_t.t.6f36_rm _____ 26921.1 __________ 1 .53.7.‘.i.-___-_.<_9-_999.1 ...........
MR1-Model #4 Dist to Den - 0.0011 156.82 <0.0001 1339.38
Prey Value +0.0524 19.56 <0.0001
Dist to Stream - 0.0016 16.02 <0.0001
Habitat 2.61 0.2711
Shrubland +0.1403 0.56 0.4536
Riparian F. - 0.2927 2.36 0.1248

 

180

Table A8. Logistic regression results showing all models within 2.0 AIC points of

the lowest AIC model for the second Mara River modeling period (MR2) based

on hyena locations collected during grazing hours (0900—1800 h). The final

selected model is in bold.

 

Model Variable Estimate Wald Stat. P AIC
MR2-Model #1 Dist to Den - 0.0010 76.81 <0.0001 982.09
Habitat 25.93 <0.0001
Shrubland +0.7639 22.74 <0.0001
Riparian F. - 0.1866 0.97 0.3251
Prey Value +0.0250 10.56 0.0012
Dist to Stream - 0.0019 7.91 0.0049
Dist to Cover - 0.0015 7.71 0.0055
MR2-Model #2 Dist to Den - 0.0010 71.10 <0.0001 983.57
Habitat 26.17 <0.0001
Shrubland +0.7678 22.76 <0.0001
Riparian F. - 0.1531 0.61 0.4338
Dist to Stream - 0.0021 8.42 0.0037
Dist to Cover - 0.0016 8.15 0.0043
Prey Value +0.0214 5.51 0.0190

 

181

 

Table A9. Logistic regression results showing all models within 2.0 AIC points of

the lowest AIC model for the second Mara River modeling period (MR2) based

on hyena locations collected during livestock-free hours (1900-0800 h). The final

selected model is in bold.

 

Model Variable Estimate Wald Stat. P AIC
MR2-Model #1 Dist to Den - 0.0006 99.95 <0.0001 1922.18

Dist to Cover - 0.0012 15.55 <0.0001

Dist to Stream - 0.0017 15.47 <0.0001

Prey Value +0.0171 6.95 0.0084

Dist to Perm +0.0003 4.04 0.0443
MR2-Model #2 Dist to Den - 0.0006 103.32 <0.0001 1924.05

Prey Value +0.0238 18.53 <0.0001

Dist to Cover - 0.0010 12.28 0.0005

Dist to Stream - 0.0014 12.00 0.0005

 

182

Table A.10. Logistic regression results showing all models within 2.0 AIC points

of the lowest AIC model, as well as the final selected model (in bold), for the third

Mara River modeling period (MR3). Models are based on hyena locations

collected during grazing hours (0900—1800 h).

 

Model Variable Estimate Wald Stat. P AIC
MR3-Model #1 Dist to Den - 0.0014 238.75 <0.0001 1782.50
Dist to Stream - 0.0056 167.97 <0.0001
Dist to Perm +0.0018 107.15 <0.0001
Habitat 12.01 0.0025
Shrubland +0.2840 4.28 0.0386
Riparian F. +0.1254 0.59 0.4407
________-____-____Eir.e_z_v_a_ly_6 ______ 69.2.96. __________ 6 .91 _________ 6 62.6.2. ...........
MR3-Model #2 Dist to Den - 0.0014 235.70 <0.0001 1784.41
Dist to Stream - 0.0055 116.05 <0.0001
Dist to Perm +0.0018 103.16 <0.0001
Habitat 10.38 0.0056
Shrub/and +0.2771 3.97 0.0462
Riparian F. +0.1206 0.55 0.4602
Prey Value +0.0194 3.80 0.0514
Dist to Cover - 0.0001 0.09 0.7600
fi'ééliia‘aéi'éé'"6:675:35? """" 16.6632 """"" {33.6i""""<'6.'6061""1'7'é2§.25'
Dist to Stream - 0.0060 231.23 <0.0001
Dist to Perm +0.0019 125.94 <0.0001
Habitat 11.18 0.0037
Shrubland +0.2694 3.90 0.0483
Riparian F. +0.1266 0.61 0.4356

 

183

Table A.11. Logistic regression results showing all models within 2.0 NO points
of the lowest AlC model for the third Mara River modeling period (MR3) based on
hyena locations collected during livestock-free hours (1900-0800 h). Only one

model was identified.

 

Model Variable Estimate Wald Stat. P AIC
MR3-Model #1 Dist to Den - 0.0008 353.32 <0.0001 3306.75
Dist to Cover - 0.0017 33.33 <0.0001
Dist to Perm +0.0006 31.52 <0.0001
Prey Value +0.0325 26.16 <0.0001
Habitat 19.59 <0.0001
Shrubland - 0.0811 0.67 0.4123
Riparian F. +0.3711 15.89 <0.0001
Dist to Stream - 0.0008 7.13 0.0076

 

184

Table A.12. Logistic regression results showing all models within 2.0 AIC points

of the lowest AIC model, as well as the final selected model (in bold), for the first

Talek West modeling period (TKW1). Models are based on hyena locations

collected during grazing hours (0900—1800 h).

 

Model Variable Estimate Wald Stat. P AIC
TKW1-Model #1 Habitat 58.97 <0.0001 1041.97
Shrubland +0.9545 25.72 <0.0001
Riparian F. +0.2456 0.64 0.4243
Dist to Den - 0.0012 32.78 <0.0001
Dist to Stream - 0.0013 8.24 0.0041
Prey Value - 0.0535 8.13 0.0044
Dist to Perm +0.0006 7.39 0.0066
LUV - 0.0206 3.47 0.0623
____________________ Pj§1.t9_§291€[_____.’:9.99.1.1--_----_.2.-.3.§_-_---.Q-_1_Z§§-----------
TKW1-Model #2 Habitat 82.76 <0.0001 1042.78
Shrub/and +0.84% 23.98 <0.0001
Riparian F. +0.1888 0.39 0.5340
Dist to Den - 0.0012 31.69 <0.0001
Prey Value - 0.0569 9.12 0.0025
Dist to Perm +0.0006 7.13 0.0076
Dist to Stream - 0.0011 6.81 0.0091
____________________ Luv0622639666469
TKW1-Model #3 Habitat 57.63 <0.0001 1043.69
Shrubland +0.8679 22.55 <0.0001
Riparian F. +0.3812 1.61 0.2047
Dist to Den - 0.0011 30.27 <0.0001
Dist to Perm +0.0007 9.59 0.0020
Prey Value - 0.0488 6.61 0.0101
Dist to Stream - 0.0011 6.54 0.0106
____________________ 6161166916.-_-__19.99.12---_-__-.2...66_-_-__.6-_666_1._________-_
TKW1-Model #4 Habitat 79.73 <0.0001 1044.52
Shrub/and +0.7471 20.12 <0.0001
Riparian F. +0.3288 1.22 0.2687
Dist to Den - 0.0010 28.68 <0.0001
Dist to Perm +0.0007 9.26 0.0023
Prey Value - 0.0526 7.55 0.0060
Dist to Stream - 0.0009 4.86 0.0274

 

185

Table A.13. Logistic regression results showing all models within 2.0 AIC points

of the lowest AIC model, as well as the final selected model (in bold), for the first

Talek West modeling period (TKW1). Models are based on hyena locations

collected during livestock-free hours (1900-0800 h).

 

Model Variable Estimate Wald Stat. P AIC
TKW1-Model #1 Dist to Den - 0.0013 87.34 <0.0001 1995.57
Habitat 21.96 <0.0001
Shrubland +0.3464 6.60 0.0102
Riparian F. +0.0131 0.00 0.9532
Prey Value - 0.0387 10.35 0.0013
Dist to Stream +0.0007 7.23 0.0072
LUV - 0.0141 3.65 0.0562
Dist to Perm +0.0003 2.38 0.1231
TRWi-"M’éaé'l’rité'"D'iéi'tb'bé'n' """" - ’0T00'1'2 """ ié'ifcié'"'"<’0.'00'01'"i§9"éf0'1'
Habitat 20.32 <0.0001
Shrubland +0.3546 7.06 0.0079
Riparian F. - 0.0369 0.03 0.8658
Prey Value - 0.0462 17.99 <0.0001
Dist to Stream +0.0009 12.18 0.0005
LUV - 0.0159 4.86 0.0275
'kaiIiraaa'rrr'a'"0131365 """"" - '0T00'1' é"""'§7‘.'2’é"""<'0 0001' "1.9—97.41-
Habitat 20.46 <0.0001
Shrubland +0.2763 4.51 0.0338
Riparian F. +0.1171 0.29 0.5896
Dist to Stream +0.0009 12.52 0.0004
Prey Value - 0.0345 8.35 0.0039
Dist to Perm +0.0003 3.70 0.0545

186

Table A.13. (cont’d).

 

Model Variable Estimate Wald Stat. P AIC
TKW1-Model #4 Dist to Den - 0.0013 87.78 <0.0001 1997.42
Habitat 17.83 0.0001
Shrubland +0.3610 6.63 0.0100
Riparian F. +0.0174 0.01 0.9378
Prey Value - 0.0387 10.39 0.0013
Dist to Stream +0.0007 6.65 0.0099
LUV - 0.0138 3.43 0.0642
Dist to Perm +0.0003 2.23 0.1356
____________________ PEEJRQSJYEL--___t9;99.92_-____--_Q-_1_§_-_____9-_7_9_22___-__-----
TKW1-Model #7 Dist to Den - 0.0011 151.17 <0.0001 1999.25
Habitat 18.19 0.0001
Shrubland +0.2800 4.70 0.0302
Riparian F. +0.0680 0.10 0.7500
Dist to Stream +0.0011 24.20 <0.0001
Prey Value - 0.0434 15.93 0.0001

 

187

Table A.14. Logistic regression results showing all models within 2.0 NO points

of the lowest AIC model, as well as the final selected model (in bold), for the

second Talek West modeling period (TKW2). Models are based on hyena

locations collected during grazing hours (0900-1800 h).

 

Model Variable Estimate Wald Stat. P AIC
TKW2-Model #1 Habitat 59.91 <0.0001 679.00
Shrubland +1 .4724 16.46 <0.0001
Riparian F. - 0.6361 0.85 0.3568
Prey Value +0.0569 27.09 <0.0001
Dist to Stream - 0.0025 12.91 0.0003
____________________ 6163362651--___--.-.9_-99_6§.---_-___6_-_69----___6-_919.6--_-_--___-
TKW2-Model #2 Habitat 59.30 <0.0001 679.78
Shrubland +1 .4558 16.04 <0.0001
Riparian F. - 0.6092 0.78 0.3777
Prey Value +0.0552 24.53 <0.0001
Dist to Stream - 0.0025 12.40 0.0004
Dist to Den - 0.0003 2.78 0.0956
____________________ Luvooos412602623
TKW2-Model #3 Habitat 45.17 <0.0001 679.88
Shrubland +1.5525 17.37 <0.0001
Riparian F. - 0.5576 0.64 0.4225
Prey Value +0.0594 27.49 <0.0001
Dist to Stream - 0.0028 13.23 0.0003
Dist to Den - 0.0004 7.18 0.0074
____________________ PJ§LLQQQY§L__“$999.11---"-___1_-_2.2__-_--_Q-_2.§.3.§-----_-..__-
TKW2-Model #4 Habitat 76.43 <0.0001 680.57
Shrubland +1.4458 15.83 <0.0001
Riparian F. - 0.2912 0.18 0.6714
Prey Value +0.0591 29.58 <0.0001
Dist to Den - 0.0002 2.59 0.1074
____________________ Luv0006619601616
TKW2-Model #5 Habitat 44.87 <0.0001 680.67
Shrubland +1.5344 16.93 <0.0001
Riparian F. - 0.5310 0.58 0.4453
Prey Value +0.0577 25.01 <0.0001
Dist to Stream - 0.0027 12.75 0.0004
Dist to Den - 0.0003 3.22 0.0728
LUV +0.0054 1.24 0.2645
Dist to Cover +0.0016 1.21 0.2709

188

Table A.14. (cont’d).

 

Model Variable Estimate Wald Stat. P AIC
TKW2-Model #6 Habitat 63.40 <0.0001 680.77
Shrubland +1.4999 17.06 <0.0001
Riparian F. - 0.6269 0.83 0.3636
Prey Value +0.0591 28.21 <0.0001
Dist to Stream - 0.0022 9.07 0.0026
Dist to Perm - 0.0004 4.46 _______ Q .0346 ___________
TRWZIM'éa’eTfli'"Habitat """""""""""""""" 6178'? <0.0001 680.77
Shrubland +1.4704 16.30 <0.0001
Riparian F. - 0.5922 0.74 0.3913
Prey Value +0.0562 24.26 <0.0001
Dist to Stream - 0.0022 9.56 0.0020
LUV +0.0067 2.07 0.1502
____________________ @6119.Ram-____.-_9_99_<22___-____1.7.6---_--_6..1.66.7.-_-__-____-
TKW2-Model #8 Habitat 59.97 <0.0001 680.80
Shrubland +1.4829 16.64 <0.0001
Riparian F. - 0.6537 0.90 0.3441
Prey Value +0.0578 26.76 <0.0001
Dist to Stream - 0.0024 10.69 0.0011
Dist to Den - 0.0003 1.89 0.1695
____________________ Pj§!_t9.E§IEU--__-_:_9_99.Ql--_-___-_Q-_2_9_-_--__Q-.§§.5_4--_-_--_-__
TKW2-Model #15 Habitat 66.29 <0.0001 683.60
Shrubland +1.4369 15.71 <0.0001
Riparian F. - 0.4341 0.40 0.5266
Prey Value +0.0546 27.32 <0.0001
Dist to Stream - 0.0025 12.39 0.0004

 

189

Table A.15. Logistic regression results showing all models within 2.0 AIC points

of the lowest AIC model for the second Talek West modeling period (TKW2)

based on hyena locations collected during livestock-free hours (1900—0800 h).

The final selected model is in bold.

 

Model Variable Estimate Wald Stat. P AIC
TKW2-Model #1 Prey Value +0.0866 91.80 <0.0001 1168.82
Habitat 53.96 <0.0001
Shrubland +0.4893 10.63 0.0011
Riparian F. +0.3499 1.94 0.1640
Dist to Den - 0.0013 49.36 <0.0001
Dist to Stream - 0.0020 14.96 0.0001
Dist to Perm +0.0004 3.32 0.0685
'T'ki/Viiiiri'éaé'i'ri'zmi'a'ra'y’ vait'ié """"" J 013236?”’0’7’.‘2'i"""<'0 0001' "T176023"
Dist to Den - 0.0013 50.15 <0.0001
Habitat 43.62 <0.0001
Shrubland +0.5207 1 1.28 0.0008
Riparian F. +0.3859 2.28 0.1309
Dist to Stream - 0.0021 14.96 0.0001
Dist to Perm +0.0004 3.33 0.0682
Dist to Cover +0.0009 0.79 0.3739
’T'kWiIIiriéa'éi 556%; "value; """"" «5 00562"'"""'9'5'.'9’i"""<'0.'000'1’ "i’1’7’0i24'
Dist to Den - 0.0011 87.58 <0.0001
Habitat 54.82 <0.0001
Shrubland +0.5155 12.19 0.0004
Riparian F. +0.3117 1.59 0.2070
Dist to Stream - 0.0016 11.44 0.0007
'T’kWi-‘rviéaél'idimfi'ré'y Véifié """"" «5 0555'?"m"8’7'79'5"""<’6T000'1'"T1'7'0'.3§
Habitat 52.98 <0.0001
Shrubland +0.4767 9.92 0.0016
Riparian F. +0.3646 2.09 0.1486
Dist to Den - 0.0013 46.98 <0.0001
Dist to Stream - 0.0020 14.65 0.0001
Dist to Perm +0.0005 3.66 0.0556
LUV +0.0023 0.48 0.4906

 

190

Table A.16. Logistic regression results showing all models within 2.0 AIC points

of the lowest AIC model for the third Talek West modeling period (TKW3) based

on hyena locations collected during grazing hours (0900—1800 h). The final

selected model is in bold.

 

Model Variable Estimate Wald Stat. P AlC
TKW3-Model #1 Dist to Den - 0.0009 33.32 <0.0001 512.76
Habitat 23.75 <0.0001
Shrubland +1 .0656 23.75 <0.0001
Dist to Cover - 0.0043 4.68 0.0305
TKWfi-"M'éc'i'e'l’ri'zmb'ié't'tb' 'ESé'n' """" - '0700'09'"""3539"”207000711436"
Habitat 23.95 <0.0001
Shrubland +1.0723 23.95 <0.0001
Dist to Cover - 0.0044 4.87 0.0273
LUV - 0.0064 0.55 0.4571
'T'kWé—"Iiriéaéi'ri's'"Fi'a'Bit‘at""""'"'"""'"""""2'3'.'60"""<'0.'0001'"5'1'4'T45"
Shrubland +1.0686 23.80 <0.0001
Dist to Den - 0.0008 17.33 <0.0001
Dist to Cover - 0.0044 4.88 0.0272
____________________ PJ§1.t9..S.t!_e_a.[U-_--.t9.99.Q§-_---__-.Q-.3.1--_-_-.Q-_5.Z§.2_-_____-__.--
TKW3-Model #4 Dist to Den - 0.0009 32.94 <0.0001 514.57
Habitat 23.84 <0.0001
Shrubland +1.0689 23.84 <0.0001
Dist to Cover - 0.0044 4.82 0.0281
Dist to Perm +0.0001 0.19 0.6594
kaéI/i'aaai'rié"136113636 """"" - ”03666 """" 66.46"""'<'6.'600'1’"5'1'4Té§'
Habitat 23.76 <0.0001
Shrubland +1 .0667 23.76 <0.0001
Dist to Cover - 0.0044 4.78 0.0288
Prey Value - 0.0102 0.14 0.7074

 

191

Table A.17. Logistic regression results showing all models within 2.0 AIC points

of the lowest AIC model for the third Talek West modeling period (TKW3) based

on hyena locations collected during livestock-free hours (1900—0800 h). The final

selected model is in bold.

 

Model Variable Estimate Wald Stat. P AIC
TKW3-Model #1 Dist to Perm - 0.0010 29.13 <0.0001 930.82

Dist to Den - 0.0004 10.57 0.0011

Dist to Stream +0.0014 7.62 0.0058

Dist to Cover - 0.0016 5.53 0.0187

Prey Value +0.0358 3.49 0.0616
flagrant;“sigrg‘sgg; """" - 33676 """"" 25.82-"“-<-0.-000-1"-93-1.-55“

Dist to Den - 0.0004 9.42 0.0021

Dist to Stream +0.0015 8.22 0.0041

Dist to Cover - 0.0015 4.90 0.0268

Prey Value +0.03% 4.20 0.0404
____________________ aux-_-________--_:9;99_7.6_-_-_--136--_--.9-_2.69.7__-___-___--
TKW3-Model #3 Dist to Perm - 0.0011 38.88 <0.0001 932.13

Dist to Den - 0.0005 12.10 0.0005

Dist to Stream +0.0016 9.95 0.0016

Dist to Cover - 0.0017 6.26 0.0123

 

192

Table A.18. Logistic regression results showing all models within 2.0 NO points
of the lowest AIC model for the fourth Talek West modeling period (TKW4).
Using only hyena locations collected during grazing hours (0900—1800 h), only

one model was identified.

Model Variable Estimate Wald Stat. P AIC

 

TKW4-Model #1 Habitat 75.93 <0.0001 —

Shrubland +1.6213 75.93 <0.0001

Prey Value +0.0553 29.90 <0.0001
Dist to Den - 0.0005 14.48 0.0001
Dist to Cover - 0.0028 12.58 0.0004
Dist to Perm - 0.0006 8.78 0.0030
LUV +0.0139 4.25 0.0392

 

193

Table A.19. Logistic regression results showing all models within 2.0 AIC points

of the lowest AIC model for the fourth Talek West modeling period (TKW4) based

on hyena locations collected during livestock-free hours (1900—0800 h). The final

selected model is in bold.

 

Model Variable Estimate Wald Stat. P AIC
TKW4-Model #1 Habitat 45.04 <0.0001 1469.91
Shrubland +1 .3926 28.96 <0.0001
Riparian F. - 1.6321 11.40 0.0007
Dist to Perm - 0.0009 38.71 <0.0001
Dist to Den - 0.0007 38.43 <0.0001
Dist to Cover +0.0026 21.31 <0.0001
Prey Value - 0.0201 6.26 0.0124
Dist to Stream - 0.0011 6.00 0.0143
TKW4-Model #2 Habitat 45.46 <0.0001 1470.57
Shrubland +1 .4036 29.37 <0.0001
Riparian F. — 1.6521 11.66 0.0006
Dist to Perm - 0.0009 39.77 <0.0001
Dist to Den - 0.0007 38.05 <0.0001
Dist to Cover +0.0026 21.02 <0.0001
Prey Value - 0.0221 7.15 0.0075
Dist to Stream - 0.0011 5.76 0.0164
LUV - 0.0072 1.22 0.2686

 

194

Table A.20. Results from all logistic regression models considered for the
prediction of the probability of hyena attack on livestock within Maasai bomas
(villages). Results shown are from all models within 2.0 AlCc points of the

optimal model (considered Model #1, also shown in Table 4.3).

Estimate Wald Stat. p-value AlCc

 

Model #1 69.74
Boma houses (#) +0.39553 13.5264 0.0002
Distance to Reserve (m) - 0.00164 3.9226 0.0476
12133;} '43 """""""""""""""""""""""""""""""""""" % Effie
Boma houses (#) +0.4300 13.3083 0.0003
Distance to Reserve (m) - 0.0019 4.3936 0.0361
Bomas within 200 m (#) +0.2033 1.5798 0.2088
Model #3 71.64
Boma houses (#) +0.4066 13.2797 0.0003
Distance to Reserve - 0.0017 3.8971 0.0484
Distance to closest boma (m) - 0.0007 0.2605 0.6098

 

195

Table A.21. Results from all logistic regression models considered for the

prediction of the probability of leopard attack on livestock within Maasai bomas

(villages). Results shown are from all models within 2.0 AlCc points of the

optimal model (considered Model #1, also shown in Table 4.3).

 

Estimate Wald Stat. p-value AlCc
Model #1 68.28
Bomas within 200 m (#) - 0.7476 7.3760 0.0066
#3555 """""""""""""""""""""""""""""""""""""" a 3.40
Bomas within 200 m (#) - 0.8845 7.8145 0.0052
Distance to Reserve (m) - 0.0011 2.1717 0.1406
0.35;. """""""""""""""""""""""""""""""""""" é 3.37
Bomas within 200m (#) - 0.7262 6.8753 0.0087
Boma houses (#) +0.0650 1.0791 0.2989
Model #4 70.34
Bomas within 200m (#) - 0.8455 7.1457 0.0075
Boma houses (#) +0.0561 0.7316 0.3924
PE‘E’]?EE9-‘3§§?IYE-_-_______:-9;99.1_9 ________ 1 £199. _________ 9 .1191 ____________
Model #5 70.35
Bomas within 200 m (#) - 0.7262 7.0435 0.0080
Distance to bush (m) - 0.0004 0.1457 0.7027

 

196

Table A.22. Results from all logistic regression models considered for the
prediction of the probability of hyena attack on livestock within sheep/goat (shoat)
enclosures. Results shown are from all models within 2.0 AlCc points of the

optimal model (considered Model #1, also shown in Table 4.4).

Estimate Wald Stat. p—value AlCc

 

Model #1 83.20
6666.89.66? _________________________ t 915.61-"---23912.---"9.9.925 ___________
Model #2 83.46
Bush fence? +0.9439 9.6430 0.0019
6.66666666663616616610. _________ t 9.9.9159._______2_-9.§9_5.-__--9.-9§é§_ __________
Model #3 84.14
Bush Fence? +0.8825 9.3716 0.0022

999.3. .61 swear. it) ___________________ t 9.10.1.1---__-9_-.7_§§9_--_--9.-§_§99. __________
Model #4 84.47
Bush fence? +0.8122 6.6553 0.0099

Sheep and goats inside (#) +0.0037 2.2622 0.1326
56666661686618). ________________ - 9.13125 ________ 1 _-§_1§9____-_9.-.2.5.1_3. __________
Model #5 84.84
Bush fence? +0.7148 5.9064 0.0151
H9P.§?§.9f_9!"fl?!.(fi). ________________ - 6.9.9.46.---_6_-6_666______9.-1669 ___________
Model #6 84.97
Bush fence? +0.7686 6.4912 0.0108

Dogs of owner (#) +0.1768 1.7483 0.1861
E9P.§?§_9f.°_!!'1?f-(fl ________________ - 6.1.669 ________ 1 _-_2_669.-____9_-_2_699 ___________
Model #7 85.07
Bush Fence? +0.9915 10.0447 0.0015

Sheep and Goats inside (#) +0.0040 2.9951 0.0835
P°9_§2[9YYPEES?) ___________________ t 9.- 9166_______.6-_3.6?6______9.66166 ___________
Model #8 85.20
Bush fence? +0.9182 8.3833 0.0038

Distance to closest enclosure (m) +0.0007 0.1803 0.6711

 

197

Table A.23. Results from all logistic regression models considered for the
prediction of the probability of leopard attack on livestock within sheep/goat
(shoat) enclosures. Results shown are from all models within 2.0 AlCc points of

the optimal model (considered Model #1, also shown in Table 4.4).

Estimate Wald Stat. p-value AlCc

 

Model #1 84.93
36161612662 _________________________ t 9.9192_-____1_1_.6.621___-_9_.9.996 ___________
Model #2 85.89
Pole fence? +0.9304 9.9706 0.0016
_SI‘EEBEPE 9993285186195). ......... ’5 9.- 999.8"__-_-9.-_1_§9§______9.-_7.1.§.1 ___________
Model #3 86.82
Pole Fence? +0.8825 9.3716 0.0022
Distance to closest enclosure (m) +0.0010 0.5557 0.4560

 

198

LITERATURE CITED

Altmann, J., and P. Muruthi. 1988. Differences in daily life between
semiprovisioned and wild-feeding baboons. American Journal of Primatology
15:213-221.

Andelt, W. F. 1985. Behavioral ecology of coyotes in South Texas. Wildlife
Monographs 9421-45.

Arcese, P., L. F. Keller, and J. R. Cary. 1997. Why hire a behaviorist into a
conservation or management team? Pages 48-71 in J. L. Clemmons, and R.
Buchholz, editors. Behavioral Approaches to Conservation in the Wild.
Cambridge University, Cambridge.

Arcese, P., and A. R. E. Sinclair. 1997. The role of protected areas as ecological
baselines. Journal of Wildlife Management 61 :587-602.

Avenant, N. L., and J. A. J. Nel. 1998. Home-range use, activity, and density of
caracal in relation to prey density. African Journal of Ecology 36:347-359.

Ballard, W. B., J. S. Whitman, and C. L. Gardner. 1987. Ecology of an exploited
wolf population in south-central Alaska. Wildlife Monographs 98:1-54.

Bart, J., M. A. Fligner, and W. l. Notz 1998. Sampling and Statistical Methods for
Behavioral Ecologists. Cambridge University Press.

Beckmann, J. P., and J. Berger. 2003a. Rapid ecological and behavioral changes
in carnivores: the responses of black bears (Ursus amen'canus) to altered
food. Journal of Zoology 261 :207-212.

Beckmann, J. P., and J. Berger. 2003b. Using black bears to test ideal-free
distribution models experimentally. Journal of Mammalogy 84:594-606.

Beier, P., D. Choate, and R. H. Barrett. 1995. Movement patterns of mountain
lions during different behaviors. Journal of Mammalogy 76:1056-1070.

Bekoff, M., and M. C. Wells. 1982. Behavioral ecology of coyotes: Social
organization, rearing patterns, space use, and resource defense. Zeitschrift
fur Tierpsychologie 60:281-305.

Berger, J. 1999. Anthropogenic extinction of top carnivores and interspecific
animal behaviour: implications of the rapid decoupling of a web involving
wolves, bears, moose and ravens. Proceedings of the Royal Society of
London Series B 266:2261-2267.

199

Blanchard, B. M., and R. R. Knight. 1991. Movements of Yellowstone Grizzly
Bears. Biological Conservation 58:41-67.

Bothma, J. d. P., and J. A. J. Nel. 1980. Winter food and foraging behavior of the
aardwolf (Proteles cristatus) in the Namib-Nawkluff Park. Madoqua 12:141-
147.

Boydston, E. E. 2001. Socio-ecological determinants of space utilization in the
spotted hyena, Crocuta crocuta. Page 172. Department of Zoology. Michigan
State University, East Lansing.

Boydston, E. E., K. M. Kapheim, and K. Holekamp. 2006. Patterns of den
occupation by the spotted hyena (Crocuta crocuta). African Journal of
Ecology 44:77-86.

Boydston, E. E., K. M. Kapheim, M. Szykman, and K. E. Holekamp. 2003a.
Individual variation in space utilization by female spotted hyenas. Journal of
Mammalogy 84:1006-1018.

Boydston, E. E., K. M. Kapheim, H. E. Watts, M. Szykman, and K. E. Holekamp.
2003b. Altered behaviour in spotted hyenas associated with increased human
activity. Animal Conservation 6:207-219.

Boydston, E. E., T. L. Morelli, and K. E. Holekamp. 2001. Sex differences in
territorial behavior exhibited by the spotted hyena (Hyaenidae, Crocuta
crocuta). Ethology 107:369-385.

Brashares, J. S. 2003. Behavioral, ecological, and life-history correlates of
mammal extinctions in West Africa. Conservation Biology 17:733-743.

Brashares, J. S., P. Arcese, and M. K. Sam. 2001. Human demography and
reserve size predict wildlife extinction in West Africa. Proceedings of the
Royal Society of London Series B 268:2473-2478.

Broten, M. D., and M. Y. Said. 1995. Population trends of ungulates in and
around Kenya's Masai Mara Reserve. Pages 169-193 in A. R. E. Sinclair, and
P. Arcese, editors. Serengeti ll: dynamics, management, and conservation of
an ecosystem. University of Chicago Press, Chicago.

Burnham, K. P., and D. R. Anderson 2002. Model Selection and Multimodel
Inference. A Practical lnfonriation-Theoretic Approach. Springer, New York.

Burnsilver, S., R. Boone, and K. Galvin. 2003. Linking pastoralists to a
heterogenous landscape: The case of four Maasai group ranches in Kajiado
district, Kenya. Pages 173-199 in J. Fox, V. Mishra, R. Rindfuss, and S.
Walsh, editors. People and the environment: approaches for linking

200

household and community surveys to remote sensing and GIS. Kluwer
Academic Publishing, Boston, MA.

Butler, J. R. A. 2000. The economic costs of wildlife predation on livestock in
Gokwe communal land, Zimbabwe. African Journal of Ecology 38:23-30.

Caro, T. 1999a. The behaviour-conservation interface. Trends in Ecology and
Evolution 14:266-269.

Caro, T. M. 1998. How do we refocus behavioral ecology to address
conservation issues directly? Pages 557-565 in T. M. Caro, editor. Behavioral
Ecology and Conservation Biology. Oxford University Press, New York, NY.

Caro, T. M. 1999b. Demography and behavior of African mammals subject to
exploitation. Biological Conservation 91:91-97.

Caro, T. M., and S. M. Durant. 1995. The importance of behavioral ecology for
conservation biology: examples from Serengeti carnivores in A. R. E. Sinclair,
and P. Arcese, editors. Serengeti II. University of Chicago Press, Chicago.

Caro, T. M., N. Pelkey, M. Borner, K. L. I. Campbell, B. L. Woodworth, B. P.
Farm, J. O. Kuwai, S. A. Huish, and E. L. M. Severre. 1998. Consequences of
different forms of conservation for large mammals in Tanzania: preliminary
analyses. African Journal of Ecology 36:303-320.

Chapman, C. A. 1990. Association patterns of spider monkeys: the influence of
ecology and sex on social organization. Behavioral Ecology and Sociobiology
26:409-414.

Chavez, A. S. G. E. M. 2006. Landscape use and movements of wolves in
relation to livestock in a wildland-agriculture matrix. Journal of Wildlife
Management 70:1079-1086.

Ciucci, P., and L. Boitani. 1998. Wolf and dog depredation on livestock in central
Italy. Wildlife Society Bulletin 26:504-514.

Ciucci, P., L. Boitani, F. Francisci, and G. Andreoli. 1997. Home range, activity
and movements of a wolf pack in central Italy. Journal of Zoology 2432803-
819.

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

Connor, M. M., M. M. Jaeger, T. J. Weller, and D. R. McCullough. 1998. Effect of
coyote removal on sheep depredation in northern California. Journal of
Wildlife Management 62:690-699.

201

Conover, M. 2002. Resolving human-wildlife conflicts: The science of wildlife
damage management. Lewis Publishers, Boca Raton, Florida.

Cooper, S. M. 1989. Clan sizes of spotted hyaenas in the Savuti Region of the
Chobe National Park, Botswana. Botswana Notes and Records 21:121-131.

Cooper, S. M., K. E. Holekamp, and L. Smale. 1999. A seasonal feast: long-term
analysis of feeding behaviour in the spotted hyaena Crocuta crocuta. African
Journal of Ecology 37:149-160.

Craighead, J. J., and F. C. Craighead. 1971. Grizzly bear-man relationships in
Yellowstone National Park. BioScience 21:845-857.

Crooks, K. R., and M. E. Soulé. 1999. Mesopredator release and avifaunal
extinctions in a fragmented system. Nature 400:563-566.

Daan, S. 1981. Adaptive daily strategies in behavior. Pages 275-298 in J.
Aschoff, editor. Handbook of Behavioral Neurobiology 4: Biological Rhythms.
Plenum Press, New York.

Daan, S., and J. Aschoff. 1982. Circadian contributions to survival. Pages 305-
338 in J. S. Aschoff, S. Daan, and G. A. Gross, editors. Vertebrate Circadian
Systems: Structure and physiology. Springer-Verlag, New York, USA.

Doncaster, C. P., C. R. Dickman, and D. W. Macdonald. 1990. Feeding ecology
of red foxes (Vulpes vulpes) in the city of Oxford, England. Journal of
Mammalogy 71:188-194.

Doran, D. 1997. Influence of seasonality on activity patterns, feeding behavior,
ranging, and grouping patterns in Tai chimpanzees. International Journal of
Primatology 18:183-206.

Drew, G. S., and J. A. Bissonette. 1997. Winter activity patterns of American
martens (Martes americana): Rejection of the hypothesis of thermal-cost
minimization. Canadian Journal of Zoology 75:812-816.

Dublin, H. T. 1995. Vegetation dynamics in the Serengeti-Mara ecosystem: the
role of elephants, fire and other factors. Pages 71-90 in A. R. E. Sinclair, and
P. Arcese, editors. Serengeti ll: dynamics, management and conservation of
an ecosystem. Chicago University Press, Chicago, IL.

Durant, S. M. 1998. Competition refuges and coexistence: an example from
Serengeti carnivores. Journal of Animal Ecology 67:370-386.

East, M., H. Hofer, and A. Turk. 1989. Functions of birth dens in spotted hyaenas
(Crocuta crocuta). Journal of Zoology. London 219:690-697.

202

East, M. L., T. Burke, K. Wilhelm, C. Greig, and H. Hofer. 2003. Sexual conflicts
in spotted hyenas: male and female mating tactics and their reproductive
outcome with respect to age, social status and tenure. Proceedings of the
Royal Society of London Series B 270:1247-1254.

East, M. L., and H. Hofer. 1991a. Loud calling in a female-dominated mammalian
society. I. Structure and composition of whooping bouts of spotted hyenas,
Crocuta crocuta. Animal Behaviour 42:637-649.

East, M. L., and H. Hofer. 1991b. Loud-calling in a female-dominated mammalian
society. II. Behavioural contexts and functions of whooping of spotted
hyaenas, Crocuta crocuta. Animal Behaviour 42:651-669.

East, M. L., and H. Hofer. 2001. Male spotted hyenas (Crocuta crocuta) queue
for status in social groups dominated by females. Behavioral Ecology 12:558-
568.

Ewer, R. F. 1973. The Carnivores. Cornell University Press, Ithaca.

Fedriani, J. M., T. K. Fuller, and R. M. Sauvajot. 2001. Does availability of
anthropogenic food enhance densities of omnivorous mammals? An example
with coyotes in southern California. Ecography 24:325-331.

Ferguson, J. W. H., J. S. Galpin, and M. J. De Wet. 1988. Factors affecting the
activity patterns of black-backed jackals Canis mesomelas. Journal of
Zoology 214:55-69.

Fernandez, N., and F. Palomares. 2000. The selection of breeding dens by the
endangered Iberian lynx (Lynx pardinus): implications for its conservation.
Biological Conservation 94:51-61.

Frank, L. 1998. Living with lions: carnivore conservation and livestock in Laikipia
District, Kenya. Development Alternatives, Bethesda, Maryland.

Frank, L. 2000. Spotted hyenas and livestock in Laikipia District, Kenya. I. U. C.
N. Hyaena Specialist Group Newsletter 7:13-18.

Frank, L. G. 1986a. Social organization of the spotted hyaena (Crocuta crocuta).
l. Demography. Animal Behaviour 34:1500-1509.

Frank, L. G. 1986b. Social organization of the spotted hyaena (Crocuta crocuta).
ll. Dominance and reproduction. Animal Behaviour 34:1510-1527.

Fuller, T. K. 1995. An international review of large carnviore conservation status.
Pages 410-412 in J. A. Bissonette, and P. R. Krausman, editors. Integrating

203

people and wildlife for a sustainable future. The Wildlife Society, Bethesda,
Maryland.

Fuller, T. K., and P. W. Kat. 1990. Movements, activity, and prey relationships of
African wild dogs (Lycaon pictus) near Aitong, southwestern Kenya. African
Journal of Ecology 28:330-350.

Fuller, T. K., and L. B. Keith. 1980. Wolf population dynamics and prey
relationships in northeastern Alberta. Journal of Wildlife Management 44:583-
602.

Garshelis, D. L., and M. R. Pelton. 1980. Activity of black bears in the Great
Smoky Mountains National Park. Journal of Mammalogy 6128-19.

Gasaway, W. C., K. T. Mossestad, and P. E. Stander. 1991. Food acquisition by
spotted hyaenas in Etosha National Park, Namibia: predation versus
scavenging. African Journal of Ecology 29:64-75.

Geffen, E., and D. W. Macdonald. 1993. Activity and movement patterns of
Blanford's foxes. Journal of Mammalogy 74:455-463.

Gese, E. M., O. J. Rongstad, and W. R. Mytton. 1989. Changes in coyote
movements due to military activity. Journal of Wildlife Management 53:334-
339.

Gibeau, M. L., A. P. Clevenger, S. Herrero, and J. Wierzchowski. 2002. Grizzly
bear response to human development and activities in the Bow River
Watershed, Alberta, Canada. Biological Conservation 103:227-236.

Gilchrist, J. S., and E. Otali. 2002. The effects of refuse-feeding on home-range
use, group size, and intergroup encounters in the banded mongoose.
Canadian Journal of Zoology 80:1795-1802.

Gill, J. A., and W. J. Sutherland. 2000. Predicting the consequences of human
disturbance from behavioral decisions. Pages 51-64 in L. M. Gosling, and W.
J. Sutherland, editors. Behavior and Conservation. Cambridge University
Press, Cambridge, UK.

Ginsberg, J. R., and D. W. Macdonald. 1990. Foxes, wolves, jackals and dogs:
an action plan for the conservation of canids. World Conservation Union,
Gland, Switzerland.

Graham, K., A. P. Beckerman, and S. Thirgood. 2005. Human-predator-prey

conflicts: ecological correlates, prey losses and patterns of management.
Biological Conservation 122:159-171.

204

Harcourt, A. H., S. A. Parks, and R. Woodroffe. 2001. Human density as an
influence on species/area relationships: double jeopardy for small African
reserves? Biodiversity and Conservation 10:1011-1026.

Hemson, G. 2003. The ecology and conservation of lions: Human-wildlife conflict
in semi-arid Botswana. Page 213. University of Oxford.

Henschel, J. R. 1986. The socio-ecology of a spotted hyena Crocuta cmcuta clan
in the Kruger National Park. University of Pretoria, Pretoria, South Africa.

Henschel, J. R., and J. D. Skinner. 1987. Social relationships and dispersal
patterns in a clan of spotted hyaenas Crocuta crocuta in the Kruger National
Park. South African Journal of Zoology 22:18-24.

Henschel, J. R., and J. D. Skinner. 1990. The diet of the spotted hyenas Crocuta
crocuta in Kruger National Park. African Journal of Ecology 28:69-82.

Henschel, J. R., and J. D. Skinner. 1991. Territorial behaviour by a clan of
spotted hyaenas Crocuta crocuta. Ethology 88:223-235.

Herrero, S. 1983. Social behavior of black bears at a garbage dump in Jasper
National Park. International Conference on Bear Research and Management
5:54-70.

Hidalgo-Mihart, M. G., L. Cantu-Salazar, C. A. Lopez-Gonzalez, E. C.
Fernandez, and A. Gonzales-Romero. 2004. Effect of a landfill on the home-
range and group size of coyotes (Canis latrans) in a tropical deciduous forest.
Journal of Zoology 263:55-63.

Hofer, H., and M. L. East. 1993a. The commuting system of Serengeti spotted
hyaenas: how a predator copes with migratory prey. l. Social organization.
Animal Behaviour 46:547-557.

Hofer, H., and M. L. East. 1993b. The commuting system of Serengeti spotted
hyaenas: how a predator copes with migratory prey. ll. Intrusion pressure
and commuter's space use. Animal Behaviour 46:559-574.

Holekamp, K. E., E. E. Boydston, and L. Smale. 2000. Group travel in social
carnivores. Pages 587-627 in S. Boinski, and P. Garber, editors. On the
Move: How and why animals travel in groups. University of Chicago Press,
Chicago, IL, USA.

Holekamp, K. E., S. M. Cooper, C. I. Katona, N. A. Berry, L. G. Frank, and L.

Smale. 1997a. Patterns of association among female spotted hyenas
(Crocuta crocuta). Journal of Mammalogy 78:55-74.

205

Holekamp, K. E., and L. Smale. 1990. Provisioning and food sharing by lactating
spotted hyenas, Crocuta crocuta (Mammalia: Hyaenidae). Ethology 86:191-
202.

Holekamp, K. E., and L. Smale. 1992. Human-hyaena relations in and around the
Masai Mara National Reserve, Kenya. I. U. C. N. Hyaena Specialist Group
Newsletter 5:19-20.

Holekamp, K. E., and L. Smale. 1998. Behavioral development in the spotted
hyena. BioScience 482997-1005.

Holekamp, K. E., L. Smale, R. Berg, and S. M. Cooper. 1997b. Hunting rates and
hunting success in the spotted hyena (Cmcuta crocuta). Journal of Zoology,
London 24221-15.

Holekamp, K. E., L. Smale, and M. Szykman. 1996. Rank and reproduction in the
female spotted hyaena. Journal of Reproduction and Fertility 108:229-237.

Homewood, K., E. F. Lambin, E. Coast, A. Kariuki, l. Kikula, J. Kivelia, M. Said,
S. Serneels, and M. Thompson. 2001. Long-term changes in Serengeti-Mara

wildebeest and land cover: pastoralism, population, or policies? Proceedings
of the National Academy of Sciences of the USA. 98:12544-12549.

Homewood, K., and W. Rodgers 1991. Maasailand Ecology: pastoralist
development and wildlife conservation in Ngorogoro, Tanzania. Cambridge
University Press.

Honer, O. P., B. Wachter, M. East, V. A. Runyoro, and H. Hofer. 2005. The effect
of prey abundance and foraging tactics on the population dynamics of a
social, territorial carnivore, the spotted hyena. Oikos 108:544-554.

Hooge, P. N., and B. Eichenlaub. 2000. Animal movement extension to Arcview.
ver. 2.0. US. Geological Survey, Anchorage, AK, USA.

Hopcraft, J. G. C., A. R. E. Sinclair, and C. Packer. 2005. Planning for success:
Serengeti lions seek prey accessibility rather than abundance. Journal of
Animal Ecology 74:559-566.

lkanda, D. K. 2005. A study of the spatial dynamics and human interactions of
African lions Panthera [90 (L., 1758) in the Ngorogoro Conservation Area,
Tanzania. Page 126. Department of Wildlife Ecology. University of Dar es
salaam, Dar es salaam.

IUCN. 2006. IUCN Red list of threatened species. <iucnredlist.org> Downloaded
on February 14 2007.

206

Jenness, J. 2006. Grid Tools (Jenness Enterprises) v. 1.7 (grid_tools_jen.avx)
extension for ArcView 3.x. Jenness Enterprises,
http://www.jennessent.com/arcview/grid_tools.htm.

Karani, l. W. 1994. An assessment of depredation by lions and other predators in
the group ranches adjacent to Masai Mara National reserve. Page 70.
Department of Wildlife Management. Moi University.

Kimani, K., and J. Pickard. 1998. Recent trends and implications of group ranch
sub-division and fragmentation in Kajiado District, Kenya. Geographic Journal
164:202-213.

Kitchen, A. M., E. M. Gese, and E. R. Sohauster. 2000. Changes in coyote
activity patterns due to reduced exposure to human persecution. Canadian
Journal of Zoology 78:853-857.

Knight, R. R., B. M. Blanchard, and L. L. Eberhardt. 1988. Mortality patterns and
population sinks for Yellowstone grizzly bears, 1973-1985. Wildlife Society
Bulletin 16:121-125.

Knight, R. R., and L. L. Eberhardt. 1985. Population dynamics of Yellowstone
grizzly bears. Ecology 66:323-334.

Kolowski, J. M., and K. E. Holekamp. 2006. Spatial, temporal, and physical
characteristics of livestock depredations by large carnivores along a Kenyan
reserve border. Biological Conservation 128:529-541.

Kolowski, J. M., and K. E. Holekamp. 2007. Daily patterns of activity in the
spotted hyena. Journal of Mammology (in press).

Kruuk, H. 1972. The spotted hyena: a study of predation and social behavior.
University of Chicago Press, Chicago, Illinois.

Kruuk, H. 1976. Feeding and social behaviour of the striped hyaena (Hyaena
vulgaris Desmarest). East African Wildlife Journal 14:91-111.

Kruuk, H. 1981. The effects of large carnivores on livestock and animal
husbandry in Marsabit District, Kenya. UNEP-MAB integrated project in arid
lands. lPAL technical report E-4.

Lamprey, R. H., and R. S. Reid. 2004. Expansion of human settlement in Kenya's

Maasai Mara: what future for pastoralism and wildlife? Journal of
Biogeography 31 :997-1032.

207

Lariviere, S., J. Huot, and C. Samson. 1994. Daily activity patterns of female
black bears in a morthern mixed-forest environment. Journal of Mammalogy
75:613-620.

Lariviere, S., and F. Messier. 1997. Seasonal and daily activity patterns of striped
skunks (Mephitis mephitis) in the Canadian prairies. Journal of Zoology
243:255-262.

Lehmann, J., and C. Boesch. 2004. To fission or to fusion: effects of community
size on wild chimpanzee (Pan troglodytes verus) social organisation.
Behavioral Ecology and Sociobiology 56:207-216.

Lima, S. L., and L. M. Dill. 1990. Behavioral decisions made under the risk of
predation: a review and prospectus. Canadian Journal of Zoology 68:619-
640.

Linnell, J. D. C., J. Odden, M. E. Smith, R. Aanes, and J. E. Swenson. 1999.
Large carnivores that kill livestock: do "problem individuals" really exist?
Wildlife Society Bulletin 27:698-705.

Litvaitis, J. A., J. A. Sherburne, and J. A. Bissonette. 1986. Bobcat habitat use
and home range size in relation to prey density. Journal of Wildlife
Management 50:110-117.

Lode, T. 1995. Activity pattern of polecats Mustela puton'us L. in relation to food
habits and prey activity. Ethology 100:295-308.

Long, J. S. 1997. Regression models for categorical and limited independent
variables. Sage Publications, Thousand Oaks, CA.

Lucherini, M., and G. Crema. 1994. Seasonal variation in diet and trophic niche
of the red fox in an alpine habitat. Zeitschrift Fur Saugetierkunde-lnternational
Journal of Mammalian Biology 5921-8.

Lucherini, M., S. Lovari, and G. Crema. 1995. Habitat use and ranging behavior
of the red fox (Vulpes vulpes) in a Mediterranean rural area: ls shelter
availability a key factor? Journal of Zoology 237:577-591.

Lunn, N. J., and I. Stirling. 1985. The significance of supplemental food to polar
bears during the ice-free period of Hudson Bay. Canadian Journal of Zoology
63:2291-2297.

Machutchon, A. G., S. Himmer, H. Davis, and M. Gallagher. 1998. Temporal and
spatial activity patterns among coastal bear populations. Ursus 10:539-546.

208

Madhusudan, M. D. 2003. Living amidst large wildlife: Livestock and crop
depredation by large mammals in the interior villages of Bhadra Tiger
Reserve, South India. Environmental Management 31:466-475.

Manly, B. F. J., L. L. McDonald, D. L. Thomas, T. L. McDonald, and W. P.
Erickson 2002. Resource selection by animals: statistical design and analysis
for field studies. Kluwer Academic Publishers, Norwell, MA.

Mattson, D. J. 1990. Human impacts on bear habitat use. International
Conference on Bear Research and Management 8:33-56.

McClennen, N., R. R. Wigglesworth, and S. H. Anderson. 2001. The effect of
suburban and agricultural development on the activity patterns of coyotes
(Canis latrans). American Midland Naturalist 146:27-36.

McLellan, B. N., and D. M. Shackleton. 1988. Grizzly bears and resource-
extraction industries: Effects of roads on behaviour, habitat use and
demography. Journal of Applied Ecology 25:451-460.

Mech, D. L., E. K. Harper, T. Meier, and W. J. Paul. 2000. Asessing factors that
may predispose Minnesota farms to wolf depredations on cattle. Wildlife
Society Bulletin 28:623-629.

Meriggi, A., and S. Lovari. 1996. A review of wolf predation in southern Europe:
does the wolf prefer wild prey to livestock? Journal of Applied Ecology
33:1561-1571.

Michalski, F., R. L. P. Boulhosa, A. Faria, and C. A. Peres. 2006. Human-wildlife
conflicts in a fragmented Amazonian forest landscape: determinants of large
felid depredation on livestock. Animal Conservation 92179-188.

Mills, M. G. L. 1984. The comparative behavioural ecology of the brown hyaena
Hyaena brunnea and the spotted hyaena Crocuta crocuta in the southern
Kalahari. Koedoe 27:237-247.

Mills, M. G. L. 1987. Behavioural adaptations of brown and spotted hyaenas in
the southern Kalahari. South African Journal of Science 83:595-598.

Mills, M. G. L. 1990. Kalahari hyaenas: comparative behavioral ecology of two
species. Unwin Hyman, London.

Mills, M. G. L. 1991. Conservation management of large carnivores in Africa.
Koedoe 34:81-90.

209

Mills, M. G. L., and M. L. Gonnan. 1987. The scent-marking behaviour of the
spotted hyaena Crocuta crocuta in the southern Kalahari. Journal of Zoology
212:483-497.

Mills, M. G. L., and H. Hofer. 1998. Hyaenas: status survey and conservation
action plan. lUCN/SSC Hyaena Specialist Group. IUCN, Gland, Switzerland.

Mills, S. L., and F. F. Knowlton. 1991. Coyote space use in relation to prey
abundance. Canadian Journal of Zoology 69:1516-1521.

Mishra, C. 1997. Livestock depredation by large carnivores in the Indian trans-
Himalaya: conflict perceptions and conservation prospects. Environmental
Conservation 24:338-343.

Mizutani, F. 1999. Impact of leopards on a working ranch in Laikipia, Kenya.
African Journal of Ecology 37:211-225.

Muriuki, G., and M. Mulama. 1997. A dry total count of elephant, buffalo and key
livestock in the Masai Mara National Reserve and the adjacent areas.
Biodiversity Department Ecological Monitoring, Kenya Wildlife Service,
Nairobi, Kenya.

Muriuki, G., P. Omondi, and R. Sinange. 2000. Elephant, buffalo, giraffe and key
livestock dry season total aerial count of Masai Mara National Reserve and
the adjacent pastoral area. Research and Planning Department, Kenya
Wildlife Service, Nairobi, Kenya.

Norton-Griffiths, M. 1995. Economic incentives to develop the rangelands of the
Serengeti: Implications for wildlife conservation. Pages 588-604 in A. R. E.
Sinclair, and P. Arcese, editors. Serengeti ll: Dynamics, Management, and
Conservation of an Ecosystem. University of Chicago Press, Chicago.

Nowell, K., and P. Jackson, editors. 1996. Wild cats: status survey and
conservation action plan. IUCN/SSC Cat Specialist Group, Gland,
Switzerland.

Ogada, M. O., R. Woodroffe, N. O. Oguge, and L. Frank. 2003. Limiting
depredation by African carnivores: the role of livestock husbandry.
Conservation Biology 17:1521-1530.

Ogutu, J. O., N. Bhola, and R. Reid. 2005. The effects of pastoralism and

protection on the density and distribution of carnivores and their prey in the
Mara ecosystem of Kenya. Journal of Zoology 265:281-293.

210

Ogutu, J. O., and H. T. Dublin. 1998. The response of lions and spotted hyaenas
to sound playbacks as a technique for estimating population size. African
Journal of Ecology 36:83-95.

Ogutu, J. O., and H. T. Dublin. 2002. Demography of lions in relation to prey and
habitat in the Maasai Mara National Reserve, Kenya. African Journal of
Ecology 40:120-129.

Olson, T. L., R. C. Squibb, and B. K. Gilbert. 1998. Brown bear diurnal activity
and human use: a comparison of two salmon streams. Ursus 10:547-555.

Olson, T. L. G. B. K. 1994. Variable impacts of people on brown bear use of an
Alaskan river. International Conference on Bear Research and Management
9297-106.

Ottichilo, W. K., J. de Leeuw, A. K. Skidmore, H. H. T. Prins, and M. Y. Said.
2000. Population trends of large non-migratory wild herbivores and livestock
in the Masai Mara ecosystem, Kenya, between 1977 and 1997. African
Journal of Ecology 38:202-216.

Paragi, T. F., S. M. Arthur, and W. B. Krohn. 1994. Seasonal and circadian
activity patterns of female fishers, Martes pennanti, with kits. The Canadian
Field-Naturalist 108:52-57.

Patterson, B. D., S. M. Kasiki, E. Selempo, and R. W. Kays. 2004. Livestock
depredation by lions (Panthera Ieo) and other carnivores on ranches
neighboring Tsavo National Parks, Kenya. Biological Conservation 119:507-
516.

Polisar, J., l. Maxit, D. Scognamillo, L. Farrell, M. E. Sunquist, and J. F.
Eisenberg. 2003. Jaguars, pumas, their prey base, and cattle ranching:
ecological interpretations of a management problem. Biological Conservation
109:297-310.

Powell, R. A. 2000. Animal home ranges and territories and home range
estimators. Pages 65-110 in L. Boitani, and T. K. Fuller, editors. Research
techniques in animal ecology: controversies and consequences. Columbia
University Press, New York.

Rainy, M., and J. Rainy. 1989. High noon on the Maasai Mara. New Scientist
124248-52.

Reid, R. S., M. Rainy, J. Ogutu, R. L. Kruska, K. Kimani, M. Nyabenge, M.

McCartney, M. Kshatriya, J. Worden, L. Ng'ang'a, J. Owuor, J. Kinoti, E.
Njuguna, C. J. Wilson, and R. H. Lamprey. 2003. People, Wildlife and

211

Livestock in the Mara Ecosystem: the Mara Count 2002. International
Livestock Research Institute, Nairobi, Kenya, Nairobi, Kenya.

Reinhart, D. P. M. D. J. 1990. Bear use of cutthroat trout spawning streams in
Yellowstone National Park. International Conference on Bear Research and
Management 82343-350.

Riley, S. P. D., R. M. Sauvajot, T. K. Fuller, E. C. York, D. A. Kamradt, C.
Bromley, and R. K. Wayne. 2003. Effects of urbanization and habitat
fragmentation on bobcats and coyotes in southern California. Conservation
Biology 17:566-576.

Ripple, W. J., E. J. Larsen, R. A. Renkin, and D. W. Smith. 2001. Trophic
cascades among wolves, elk and aspen on Yellowstone National Park's
northern range. Biological Conservation 102:227-234.

Rogers, L. L., D. W. Kuehn, A. W. Erickson, E. M. Harger, L. J. Verrne, and J. J.
Ozoga. 1976. Characteristics and management of black bears that feed in
garbage dumps, campgrounds, or residential areas. lntemational Conference
on Bear Research and Management 3:169-175.

Rudnai, J. 1979. Ecology of lions in Nairobi National Park and the adjoining
Kitengela Conservation Unit in Kena. African Journal of Ecology 17:85-95.

Rusak, B. 1981. Vertebrate Behavioral Rhythms. Pages 183-213 in J. Aschoff,
editor. Handbook of Behavioral Neurobiology 4: Biological Rhythms. Plenum
Press, New York.

Sagor, J. T., J. E. Swenson, and E. Roskaft. 1997. Compatibility of brown bear
Ursus arctos and free-ranging sheep in Nonrvay. Biological Conservation
81:91-95.

Salvador, A., and P. L. Abad. 1987. Food habits of a wolf population (Canis
lupus) in Leon province, Spain. Mammalia 51:45-52.

Schaller, G. B. 1972. The Serengeti Lion. University of Chicago Press, Chicago.

Schmidt, K. 1999. Variation in daily activity of the free-living Eurasian lynx (Lynx
lynx) in Bialowieza Primeval Forest, Poland. Journal of Zoology 249:417-425.

Seaman, D. E., and R. A. Powell. 1990. Identifying patterns and intensity of
home range use. International Conference on Bear Research and
Management 82243-249.

Seaman, D. E., and R. A. Powell. 1996. An evaluation of the accuracy of kernel
density estimators for home range analysis. Ecology 77:2075-2085.

212

Serneels, S., and E. F. Lambin. 2001. Impact of land-use changes on the
wildebeest migration in the northern part of the Serengeti-Mara ecosystem.
Journal of Biogeography 28:391-407.

Serneels, S., M. Y. Said, and E. F. Lambin. 2001. Land-cover changes around a
major East-African wildlife reserve: The Mara Ecosystem (Kenya).
International Journal of Remote Sensing 2223397.

Shimooka, Y. 2005. Sexual differences in ranging of Ate/es belzebuth belzebuth
at La Macarena, Colombia. International Journal of Primatology 26:385-406.

Sillero-Zubiri, C., and M. K. Laurenson. 2001. Interactions between carnivores
and local communities: conflict or co-exlstence? Pages 282-312 in J. L.
Gittleman, S. M. Funk, D. Macdonald, and R. K. Wayne, editors. Carnivore
Conservation. Cambridge University Press, Cambridge.

Smale, L., L. G. Frank, and K. E. Holekamp. 1993. Ontogeny of dominance in
free-living spotted hyaenas: juvenile rank relations with adult females and
immigrant males. Animal Behaviour 46:467-477.

Smale, L., S. Nunes, and K. E. Holekamp. 1997. Sexually dimorphic dispersal in
mammals: patterns, causes, and consequences. Advances in the Study of
Behavior 26:181-250.

Sokal, R. R., and F. J. Rohlf 1995. Biometry: the principles and practice of
statistics in biological research. W. H. Freeman and Company, New York.

Spencer, P. 2003. Time, space and the unknown: Maasai configurations of
power and providence. Routledge, London, England.

Spong, G. 2002. Space use in lions, Panthera Ieo, in the Selous Game Reserve:
social and ecological factors. Behavioral Ecology and Sociobiology 52:303-
307.

Stahl, P., J. M. Vandel, V. Herrenschmidt, and P. Mlgot. 2001. Predation on
livestock by an expanding reintroduced lynx population; long-term trend and
spatial variability. Journal of Applied Ecology 38:674-687.

Stander, P. 1997. The ecology of lions and conflict with people in north-eastern
Namibia. Pages 10-17. Proceedings of a Symposium on Lions and Leopards
as Game Ranch Animals, Onderstepoort, South Africa.

Stander, P. E. 1992. Cooperative hunting in lions: the role of the individual.
Behavioral Ecology and Sociobiology 29:445-454.

213

StatSoft. 2002. STATISTICA 6.1 (data analysis software system). StatSoft,
Tulsa, OK.

Stoddart, L. C., R. E. Griffiths, and F. F. Knowlton. 2001. Coyote responses to
changing jackrabbit abundance affect sheep predation. Journal of Range
Management 54:15-20.

Sunquist, M. E., and F. Sunquist. 2001. Changing landscapes: consequences for
carnivores. Pages 399-418 in J. L. Gittleman, S. M. Funk, D. Macdonald, and
R. K. Wayne, editors. Carnivore Conservation. Cambridge University Press,
UK.

Symington, M. M. 1990. Fission-Fusion Social-Organization in Ate/es and Pan.
International Journal of Primatology 11:47-61.

 

Szykman, M., A. L. Engh, R. C. Van Horn, S. M. Funk, K. T. Scribner, and K. E.
Holekamp. 2001. Association patterns among male and female spotted
hyenas (Crocuta crocuta) reflect male mate choice. Behavioral Ecology and
Sociobiology 50:231-238.

Terborgh, J. 2001. Ecological meltdown in predator-free forest fragments.
Science 294: 1 923-1 925.

Theuerkauf, J., W. Jedrzejewski, K. Schmidt, and R. Gula. 2003a.
Spatiotemporal segregation of wolves from humans in the Bialowieza Forest
(Poland). Journal of Wildlife Management 67:706-716.

Theuerkauf, J., W. Jedrzejewski, K. Schmidt, H. Okarma, l. Ruczynski, S.
Sniezko, and R. Gula. 2003b. Daily patterns and duration of wolf activity in
the Bialowieza Forest, Poland. Journal of Mammalogy 84:243-253.

Theuerkauf, J., S. Rouys, and W. Jedrzejewski. 2003c. Selection of den,
rendezvous, and resting sites by wolves in the Bialowieza Forest, Poland.
Canadian Journal of Zoology 81:163-167.

Thurber, J. M., R. O. Peterson, T. D. Drummer, and S. A. Thomasma. 1994.
Gray wolf response to refuge boundaries and roads in Alaska. Wildlife Society
Bulletin 22:61-67.

Tilson, R. T., and W. J. I. Hamilton. 1984. Social dominance and feeding patterns
of spotted hyaenas. Animal Behaviour 32:715-724.

Tilson, R. T., and J. R. Henschel. 1986. Spatial arrangement of spotted hyaena

groups in a desert environment, Namibia. African Journal of Ecology 24:173-
180.

214

Treves, A., and K. U. Karanth. 2003. Human-carnivore conflict and perspectives
on carnivore management worldwide. Conservation Biology 17:1491-1499.

Trinkel, M., P. H. Fleischmann, A. F. Steindorfer, and G. Kastberger. 2004.
Spotted hyenas (Crocuta crocuta) follow migratory prey. Seasonal expansion
of a clan territory in Etosha, Namibia. Journal of Zoology 264:125-133.

Van Dyke, F. G., R. H. Brocke, H. G. Shaw, B. B. Ackerman, T. P. Hemker, and
F. G. Lindzey. 1986. Reactions of mountain lions to logging and human
activity. Journal of Wildlife Management 50:95-102.

Van Horn, R. C., T. L. McElhinney, and K. E. Holekamp. 2003. Age estimation
and dispersal in the spotted hyena (Crocuta crocuta). Journal of Mammalogy
84:1019—1030.

Vila, C., V. Urios, and J. Castroviejo. 1995. Observations on the daily activity
patterns in the Iberian wolf. Pages 335-340 in L. N. Carbyn, S. H. Fritts, and
D. R. Seip, editors. Ecology and conservation of wolves in a changing world.
Canadian Circumpolar Institute, University of Alberta, Edmonton, Alberta.

Vos, J. 2000. Food habits and livestock depredation of two Iberian wolf packs
(Canis lupus signatus) in the north of Portugal. Journal of Zoology 251 :457-
462.

Walpole, M. J., G. G. Karanja, N. W. Sitati, and N. Leader-Williams. 2003.
Wildlife and People: Conflict and Conservation in Masai Mara, Kenya. Wildlife
and Development Series No. 14, International Institute for Environment and
Development, London.

Watts, H. E., and K. E. Holekamp. in review. Ecological determinants of survival
and reproduction in the spotted hyena. Journal of Animal Ecology.

Weaver, J. L., P. C. Paquet, and L. F. Ruggiero. 1996. Resilience and
conservation of large carnivores in the Rocky Mountains. Conservation
Biology 10:964-976.

Weber, W., and A. Rabinowitz. 1996. A global perspective on large carnivore
conservation. Conservation Biology 10:1046-1054.

White, G. C., and R. A. Garrott 1990. Analysis of Wildlife Radio-Tracking Data.
Academic Press, Inc., San Diego, California.

Woodroffe, R. 2000. Predators and people: using human densities to interpret
declines of large carnivores. Animal Conservation 3:165-173.

215

Woodroffe, R. 2001. Strategies for carnivore conservation: lessons from
contemporary extinctions. Pages 61-92 in J. L. Gittleman, R. K. Wayne, D. W.
Macdonald, and S. M. Funk, editors. Carnivore Conservation. Cambridge
University Press, Cambridge, United Kingdom.

Woodroffe, R., and L. G. Frank. 2005. Lethal control of African lions (Panthera
Ieo): local and regional population impacts. Animal Conservation 8:91-98.

Woodroffe, R., and J. R. Ginsberg. 1998. Edge effects and the extinction of
populations inside protected areas. Science 280:2126-2128.

Woodroffe, R., and J. R. Ginsberg. 2000. Ranging behaviour and vulnerability to
extinction in carnivores. Pages 125-140 in L. M. Gosling, and W. J.
Sutherland, editors. Behaviour and Conservation. Cambridge University
Press, Cambridge.

Woodroffe, R., P. Lindsey, S. Romanach, A. Stein, and S. M. K. o. Ranah. 2005.
Livestock predation by endangered African wild dogs (Lycaon pictus) in
northern Kenya. Biological Conservation 124:225-234.

Worton, B. J. 1989. Kernel methods for estimating the utilization distribution in
home-range studies. Ecology 70:164-168.

Worton, B. J. 1995. Using Monte Carlo simulation to evaluate kernel-based home
range estimators. Journal of Wildlife Management 59:794-800.

Yom-Tov, Y., S. Ashkenazi, and O. Viner. 1995. Cattle predation by the golden
jackal Canis aureus in the Golan Heights, Israel. Biological Conservation
73:19-22.

Young, 8., F., and R. L. Ruff. 1982. Population dynamics and movements of
black bears in east central Alberta. Journal of Wildlife Management 46:845-
860.

Zalewski, A. 2001. Seasonal and sexual variation in diel activity rhythms of pine
marten Martes martes in the Bialowieza National Park (Poland). Acta
Theriologica 46:295-304.

Zar, J. H. 1999. Biostatistical analysis. Prentice-Hall, Englewood Cliffs, NJ.

216

       
     

Ilililjjiji11111111in1

2