|UFDC Home||myUFDC Home | Help|
This item has the following downloads:
LAND USE AND PREY DENSITY CHANGES INT THE NAKURU WILDLIFE
IMPLICATIONS FOR CHEETAH CONSERVATION
MEREDITH MORGAN EVANS
A THESIS PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
MASTER OF SCIENCE
UNIVERSITY OF FLORIDA
Meredith Morgan Evans
This thesis is dedicated to my parents.
Thank you for all of your love and support.
I would like to thank my committee (Mel Sunquist, Madan Oli and Mike Binford)
for their advice and support in the development and completion of this project. Jane
Southworth and Theresa Burscu also gave me much-appreciated help and encouragement
in all stages.
I am extremely grateful to the members of the Cheetah Conservation Fund,
especially Mary Wykstra and Cosmas Wambui, for giving me their time and expertise,
and for allowing me to share in their research. I am also indebted to the members of the
Nakuru Wildlife Forum for allowing me access to their properties, and for their kind
hospitality and cooperation.
I gratefully acknowledge the Kenyan government for allowing me to work in their
wonderful country. I also thank the University of Florida for financial support through an
Alumni Fellowship, without which this work would not have been possible. I must also
thank Dr. Stephen Humphrey and the School of Natural Resources and Environment for
the purchase of satellite imagery.
And finally, I am extremely grateful to all of my family and friends whose
tremendous support and constant encouragement helped me get through difficult times.
This proj ect would not have been as strong as it is without their love, advice, and
TABLE OF CONTENTS
ACKNOWLEDGMENT S .............. .................... iv
LI ST OF T ABLE S .........__.. ..... .___ .............._ vii..
LIST OF FIGURES ........._.___..... .__. ..............viii...
AB STRAC T ................ .............. ix
1 INTRODUCTION ................. ...............1.......... ......
Background ................. ............... ...............2.......
Features of Susceptibility .............. ...............3.....
Effects of Prey on Carnivores............... ...............3
Human-Predator Conflicts ................. ...............5.................
Factors Affecting Cheetah Declines ................. ...............6................
Description of Problem............... ...............8
Purpose of Study ................. ...............9.................
2 STUDY DE SIGN .............. ............... 10....
Study site .............. ...............10....
Image Processing ............... .... ........... .......... .............1
Image Acquisition and Pre-Processing ................. ............. ......... .......12
Image Classification ................ ...............14.................
Habitat Types............... ...............17.
Prey Density Estimates ................. ...............20........... ....
Data Collection ................. ...............20.................
DISTANCE Analyses............... ...............25
Potential Cheetah Population Size............... ...............27..
NWF and LNNP Counts ................. ...............28................
3 RE SULT S .............. ...............3 0....
Classification .............. ...............3 0....
Land cover Change ................. ...............36.................
Prey Densities .............. ...............3 8....
Prey Density Changes Over Time .............. ...............40....
4 DI SCUS SSION ............_...... ...............43...
Cheetah Population Trends ............_...... ...............43...
Land cover Change ............... .. ......_ .... ...............45..
Direct Consequences of Landcover Change. ........_.._ .... ...._. ..........._....47
Indirect Consequences of Land cover Change ..........._..__........ ...............49
Conclusion ..........._..__....._.. ...............52....
LIST OF REFERENCE S ..............__....._.. ...............55....
BIOGRAPHICAL SKETCH .............. ...............61....
LIST OF TABLES
1 Bandwidths for Landsat 7 ETM+ and Landsat 5 TM satellite imagery ........................13
2 Regression equations used to radiometrically correct the 1986 image to the 2003
im age. ............. ...............14.....
3 DISTANCE models used in analyses .............. ...............27....
4 Accuracy results for 2003 landcover classification .............. ...............35....
5 Accuracy results for 2003 suitability classification ................. ......... ................35
6 Landcover change in the Nakuru Wildlife Conservancy, (NWC), 1986 to 2003..........37
7 Class metrics for the 1986 and 2003 classifications .............. ...............37....
8 Density estimates for prey species in grassland and bushland habitats. ........................38
9 Potential cheetah population estimates as predicted by prey biomass. ................... .......40
10 NWC wide prey density estimates and regression analyses of density changes. ........42
11 Cheetah sightings in the Nakuru Wildlife Conservancy: 2000-2002 ..........................45
LIST OF FIGURES
1 Map of the Nakuru Wildlife Conservancy, Kenya ................. ....___ ................1 1
2 Spectral profies of eight landcover classes ....._._._ ............ ......_. .........1
3 Representative images of seven land cover classes .............. ...............21....
4 2003 landcover classification ........._.. ............ ...............31...
5 1986 landcover classifieation............... .............3
6 2003 suitability classification .............. ...............33....
7 1986 suitability classification .............. ...............34....
Abstract of Thesis Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Master of Science
LAND USE AND PREY DENSITY CHANGES INT THE NAKURU WILDLIFE
CONSERVANCY, KENYA: IMPLICATIONS FOR CHEETAH CONSERVATION
Meredith Morgan Evans
Chair: Melvin Sunquist
Maj or Department: School of Natural Resources and Environment
Originally found throughout Africa outside of the Sahara and into Asia, the cheetah
has disappeared from much of its former range and is under threat in those areas where it
still exists. The current decline of cheetah populations has been attributed largely to
habitat loss and a decline in prey densities. I attempt to explain the cause of the putative
decline of the cheetah population, Acinonyx jubatus, in the Nakuru Wildlife Conservancy
(NWC), Nakuru, Kenya. I examined prey density data for the NWC and analyzed land-
use changes between 1986 and 2003 as possible correlates of the purported reduction in
the cheetah population. To analyze and quantify landcover change, three Landsat
satellite images from Path 169, Rows 60 and 61 were acquired representing the entire
study area, and were classified separately using a combination of the supervised and
unsupervised classification methods. Information on the density of prey species in
different habitat types was collected using transects, and was analyzed with the program
DISTANCE. Changes in prey density over time were determined by regressing the
average density for the whole conservancy with time. Grassland landcovers in the
conservancy were reduced by almost 16%, while bush increased almost 13%, and
marginal areas overall increased almost 15%. The biggest changes were seen in the
developed and baresoil classes, with increases of 348% and 290%, respectively.
Preferred prey were found in higher densities in grassland areas as compared to bushland,
although large and small prey showed no significant differences. Only preferred prey and
Thompson' s gazelle were shown to have declined significantly in density since 1996.
Results indicate that recent changes (1986-2003) in landcover and prey availability
within the Nakuru Wildlife Conservancy are insufficient to explain the marked decline of
cheetahs in the area. Other factors, such as high human densities in NWC and
proliferation of small scale agriculture in the surrounding areas, should be explored as
possible explanations for the cheetah population decline.
Originally found throughout Africa outside of the Sahara and into Asia, the cheetah
(Acinonyx jubatus) has disappeared from much of its former range, and is under threat in
those areas where it still exists. Little more than half the countries in Africa that once
contained cheetahs still retain populations. Of those, only 1/3 support viable populations.
The largest population can currently be found in Namibia, with about 2,000-3,000
animals. Kenya has the second largest population, with between 1,000 and 2,000
animals. The populations in Asia have been lost completely, except for a relict
population of about 200 in Iran (Marker-Kraus et al., 1996; Marker-Kraus and Kraus,
1993). Many explanations for this decline have been put forth, including loss of habitat,
decline in prey abundance, genetic homozygosity, inter-specific competition, and
persecution by people; but few have been demonstrated in field studies (Myers, 1975b;
Eaton, 1974; Caro, 1994; Gros, 1998; Marker et al., 2003).
My obj ective was to evaluate the cause of the reported cheetah population decline
in the Nakuru Wildlife Conservancy, Kenya by focusing on changes in prey densities and
landcover patterns as factors that may have resulted in habitat loss for cheetahs and their
prey. Once the cause is determined, management recommendations can be made to
mitigate or even reverse this trend. An explanation for cheetah declines and management
recommendations to address it could be applicable to other predator populations as well.
All continents except for Antarctica host populations of endangered or threatened
carnivores. The red wolf (Canzis rufus), polar bear (Ursus maritimus), and some
populations of the mountain lion (Puma concolor) are found in North America. In
Africa, most of the large cats are considered at least threatened, while the African wild
dog (Lycaon pictus) is endangered. Asia is home to the snow leopard (Panthera uncia),
dhole (Cuon alpinus), all endangered tiger (Panthera tigris) populations, and several
other carnivore species such as the panda (Ailuropoda melan2oleuca) and Himalayan
black bear (Ursus thibetanus). South America is home to the vulnerable spectacled bear
(Tremarctos ornatus) and the jaguar (Panthera onca), and several species of small felids
Large carnivores face threats from many different directions including habitat loss,
direct and indirect conflict with humans, disease, and loss of genetic diversity. These
problems are exacerbated as human populations grow and expand into new areas,
changing the landscape and pushing carnivores into smaller and smaller areas. Conflict is
increased as predators move into neighborhoods, encounter domestic animals, and
compete with humans for resources and space. A maj or cause of population declines in
carnivores is direct persecution by people. Wolves (Mech, 1995) and African wild dogs
(Frank and Woodroffe, 2001) were considered vermin, and were subj ected to government
policies to systematically remove them from their ranges, even from national parks.
Carnivores are also shot for control purposes to reduce depredation on livestock or to
protect people from possible attacks. Lions (Panthera leo) are regularly shot, and spotted
hyena clans (Crocuta crocuta) are poisoned in the Laikipia District of Kenya to protect
livestock (Frank, pers. comm.). Increased contact with humans and their pets have
affected carnivores by possibly increasing the incidence of diseases such as rabies,
distemper, and parvovirus, causing devastating losses in certain populations. The
domestic dog population adj acent to the Serengeti and Masai Mara National Parks in East
Africa was implicated in the canine distemper virus outbreak that killed 30% of the lion
population (Roelke-Parker et al., 1996). Finally, the fracturing of populations into
isolated groups because of habitat fragmentation and loss has potentially increased the
level of genetic homozygosity in some species and populations, making them less
adaptable and more vulnerable to changes in the environment.
Features of Susceptibility
Large carnivores are especially susceptible to population pressures because of their
biology and behavior. Carnivores usually maintain exclusive territories and exist
naturally at low population densities. They also have low reproductive output, with long
inter-birth intervals resulting in low recruitment rates into a population. Many species of
carnivores such as the Florida panther (Puma concolor coryi) have low genetic diversity,
making them even more vulnerable to changes in the environment and unable to adapt
(Roelke et al., 1993). Finally, carnivores often come into conflict with people over
competition for resources such as prey, livestock, and space (Sillero-Zubiri and
Laurenson, 2001). As humans affect all species, this conflict can have profound effects
on carnivores, because their populations are often regulated by the quantity and quality of
available food resources.
Effects of Prey on Carnivores
Carrying capacity has been defined by Goss-Custard and Durell (1990) and restated
by Sutherland and Anderson (1993) as "those cases where the addition of a further
individual will result in the death or emigration of another." By removing prey resources
from a system, the carrying capacity is reduced, and fewer carnivores can be supported.
The quality and quantity of prey resources available in an ecosystem can determine the
fitness of the carnivores that depend on them; and can regulate the carnivore' s density,
distribution, and home range size (Fuller and Sievert, 2001; Sunquist and Sunquist, 1989;
Krauk, 1986). A decrease in the quality or quantity of available food can have both
direct and indirect demographic effects. A lack of food can result in compromised
physical Sitness, leading to an increase in adult mortality. For example, a period of hare
scarcity in parts of Canada coincided with high levels of adult lynx mortality, due at least
in part to starvation (Poole, 1994). Low food abundance can also have indirect effects,
compromising reproductive ability and the capacity to successfully raise offspring to
independence. In San Joaquin kit foxes (Vulpes macrotis mutica), a decrease in prey
biomass resulted in a decrease in reproductive success and in the density of adult foxes
the next year (White and Ralls, 1993). Low prey densities affected the nutritional status
of wolves and consequently they produced smaller litters (Boertje and Stephenson, 1992).
Lion and cheetah mothers are both known to abandon cubs in times of food scarcity or
when they have difficulty in securing sufficient prey (Hanby et al., 1995; Caro, 1994).
Prey densities also affect the space-use pattern of carnivores, by affecting home
range configuration and territory size. Lions found in the Ngorongoro Crater of
Tanzania, where prey resources are abundant year round, live at higher densities and have
smaller ranges than their conspecifies on the Serengeti plains, where prey densities
remain low except during the migration season (Hanby et al., 1995). Coyotes (Canis
latrans)trt~r~rt~t~rtrt~ in Utah were shown to have larger territories and home-range sizes during
periods of low prey abundance (Mills and Knowlton, 1991). When home ranges expand
in response to a decrease in food or other resources, the density of carnivores found in
any given area is consequently reduced.
One of the major causes of declines in carnivore populations is conflict with
humans, resulting in both direct and indirect mortality due to exploitation, competition
for resources, and the control of problem animals. The exploitation of carnivores for
products and parts to sell on a commercial market and sport hunting has reduced some
populations to alarmingly low numbers. Spotted cats have been exploited for their pelts
while tigers and bears are killed for their bones and gall bladders, respectively, for use in
Asian medicines (Sillero-Zubiri and Laurenson, 2001). Carnivores also face competition
with humans for resources such as prey and space. In the past, humans have considered
large carnivores such as wolves and pumas as competitors for game species, and have
consequently removed them to increase game populations. For example, wolves are
controlled through harvesting in interior Alaska to increase ungulate biomass (Boertj e
and Stephenson, 1992).
Space is also an issue, as large carnivores generally have large spatial requirements.
Human population expansion is inevitably eroding the land available to carnivores,
leading to the formation of small, isolated populations with reduced opportunities for
gene flow. The conversion of natural areas to human-dominated landscapes
characterized by agriculture, housing developments, livestock ranching, and other hostile
uses of land further constrain movement and foraging. Carnivores and other wildlife
species are often actively discouraged from using land under cultivation or around human
settlement for fear of losing life, limb, or a source of income. Roads often intersect
carnivore territories, resulting in barriers to dispersal and movement, fragmentation of
habitat and mortality caused by collisions with vehicles (Smith, 1999; Sunquist and
Sunquist, 2001). As humans and their domestic animals move into carnivore territories,
increased levels of depredation can occur, resulting in the loss of carnivores killed for
control purposes. In Nepal and Kenya, snow leopard (Oli, 1994) and lion (Frank, pers.
comm.) populations are threatened due to their depredation on local livestock.
Factors Affecting Cheetah Declines
Cheetahs are as vulnerable to population decline as any other carnivore, perhaps
even more so, because of the low density at which they are normally found. In one area
of Namibia, home ranges varied from 800 km2 for males to 1,500 km2 for females
(Morsbach, 1987). In the Serengeti, lions occur at a density 3-5 times greater than
cheetahs while spotted hyenas live at densities 5-10 times greater (Laurenson et al.,
1992). Interspecific competition has been implicated in the low cheetah densities found
throughout their range. Because of this competition, cheetahs avoid areas where
competitors are found, both temporally and spatially (Durant, 1998 and 2000). Cheetahs
lose their kills to lions, hyenas and other competitors and are also killed directly by them
(Myers, 1975a; Caro, 1994). In the Serengeti, 73% of the cheetahs that die before
adulthood are killed by other carnivores, making predation the largest source of mortality
for cheetahs in this area (Laurenson, 1994). Therefore, it has been suggested that the
cheetahs' best chance for survival will be through conservation in areas outside of
national parks as national parks are often refuges for the other large carnivores (Caro,
1994; Gros, 1998). Rangelands become the next best option as long as ranchers and
other property owners can be convinced of the desirability of allowing these creatures on
Another factor adding to the cheetahs' vulnerability is their genetics. O'Brien et al.
(1986) report that of the 250 species that had been studied, "the cheetah has the least
amount of genetic variety". The lack of genetic diversity makes the cheetah susceptible
to stochasticity in the environment. While loss of genetic diversity has not been shown to
cause declines in populations on its own (Caro and Laurenson, 1994), it can be a
contributing factor if it keeps a population from adapting to a changing environment
(O'Brien et al., 1985).
The current decline of cheetah populations has been attributed largely to habitat
loss and a decline in prey densities (Caro, 1994; Marker-Kraus and Kraus, 1993; Myers,
1975b). While there have been many studies of cheetahs, few have looked at them
outside national parks. The works of Caro (1994), Durant (1998, 2000), and Kelly,
Laurenson and Fitzgibbon (1998) were done in the Serengeti. Eaton (1974) studied them
in Nairobi National Park. Other studies concentrated on captive populations or
populations under direct persecution by humans. Yet it has been stated that the cheetah's
best hope for survival lies in their conservation outside of protected areas.
Cheetahs in Kenya face the same pressures as cheetahs in other parts of their range.
Gros (1998) concluded that the population of cheetahs in Kenya has remained stable
between 1970 and 1990, but these figures are based on comparisons of densities in two
national parks. She also goes on to say that the maj ority of cheetahs in Kenya live
outside of parks, so the conclusion she reached of a stable population may not be
applicable to the maj ority of the population or the land they inhabit. The land inside
national parks is unlikely to have changed extensively due to its protected status, whereas
land outside of parks will more likely face pressures from a growing human population
for conversion to human uses. The Nakuru area has experienced extensive growth in the
last 17 years due to the growth of the flower industry and human population growth in
general (Wykstra, pers. comm.). Most likely, a variety of factors are contributing to the
decline in the cheetah population of Nakuru, but habitat loss affecting prey densities and
the cheetah directly are probably the most significant contributors. As these issues have
the potential to affect carnivores everywhere, any insight into their impacts would be
beneficial for the development of management schemes addressing habitat loss and its
Description of Problem
In 2000, members of the Nakuru Wildlife Forum (NWF), a group comprised of
private and public land owners and managers who work together to make landscape level
management decisions for the benefit of the Nakuru Wildlife Conservancy (NWC),
contacted the Cheetah Conservation Fund in Namibia to express their concern about
cheetahs in the NWC area. They had noticed a decline in the cheetah population and in
wildlife numbers in general since the early 1990s, and were wondering what the cause of
the decline could be. The Cheetah Conservation Fund (CCF) set up a satellite program in
the Nakuru area of Kenya with the primary purpose of estimating the current cheetah
population size and the cause of a decline, if it did in fact exist. The Nakuru-Naivasha
area encompassed by the conservancy has experienced phenomenal growth in the human
population with the growth of the commercial flower farm industry along Lake Naivasha.
It has also experienced a large amount of growth in developed areas due to the
infrastructure necessary to support the flower industry and the workers employed there.
The large growth in the human population combined with the growth in developed areas
are most likely having a negative impact on the wildlife found in the Nakuru Wildlife
Purpose of Study
The purpose of this study is to evaluate the two leading hypotheses for the cause of
the putative decline of the cheetah population in the Nakuru Wildlife Conservancy,
Nakuru, Kenya. Factors most likely affecting the Nakuru population include changes in
availability of their preferred prey species, and habitat loss or degradation, though other
possibilities exist. I examined prey density data for the NWC and analyzed land use
changes over the last 17 years as possible correlates of the purported reduction in the
cheetah population. With this information, steps can be taken to stop or reverse the
cheetah population decline, thus conserving them in the Nakuru area.
This study was conducted throughout and immediately surrounding the Nakuru
Wildlife Conservancy, a 350,000 acre area managed by the Nakuru Wildlife Forum and
located in the Nakuru District of Kenya, NW of Nairobi (-Oo 27' 54" latitude and 360 12'
4" longitude) (Figure 1). A variety of land uses are found in the NWC, including cattle
ranching, subsistence and commercial agriculture, flower farming for export, government
holdings and three national parks. The third largest city in Kenya, Nakuru, is found just
north of the northern end of the conservancy. Two other important towns in the area are
Gilgil and Naivasha.
The area is mostly semi-arid savanna with grassland and leleshwa (Ta~rchonanthus
camphoratus) and acacia (Acacia sp.) bushland. Forests of yellow fever acacia (Acacia
xanthophloea) are found along the three lakes and rivers. There are two rainy seasons,
the short rains fall in October and November while the long rains fall from March
through June. NWC sustains a diverse wildlife community, though many of the large and
destructive mammals have been killed or driven out of the area, including elephants and
lions. Spotted hyenas are persecuted but they have managed to maintain a small but
The local people who make up the NWC suspected a decline in the population of
cheetah and other wildlife in the last 15 years (1985-2000) and approached the Cheetah
Conservation Fund (CCF) to determine the cause. An increase in the human density has
occurred in the same time that the wildlife densities had decreased (Wykstra-Ross and
Marker, pers. comm.). A survey conducted by Gros in 1990 (Gros, 1998) supports the
idea that the cheetah population in Nakuru is declining. Currently, very little is known
about the populations of cheetahs, their prey or of competitor species in the Nakuru area.
Suden~~hp l Eth i
I ....11...1.1 |> I1 .
Figure 1. Map of the Nakuru Wildlife Conservancy, Kenya
To analyze and quantify landcover change, three Landsat satellite images from Path
169, Rows 60 and 61 were acquired representing the entire study area. The dates used for
the analysis were 28 January 1986, 06 February 1995 and 04 February 2003. Care was
taken to choose images with anniversary dates as close together as possible to minimize
differences in spectral signatures of vegetation and other landcovers due to seasonal
variation. While these dates did not fall within the time frame of Hield work, they were
chosen due to their ease of acquisition and availability. Landsat images have a pixel size
of 28.5 m x 28.5 m giving them a spatial resolution fine enough to distinguish details in
the landscape that would be seen by cheetahs and their prey. All image processing was
done in Leica Systems Erdas Imagine 8.6 unless otherwise indicated.
Image Acquisition and Pre-Processing
2003: The 2003 image is an Enhanced Thematic Mapper (1G) image (ETM+) from
Landsat 7. The 1G designation indicates that the image has been radiometrically and
geometrically corrected by USGS. The study site crosses two images but the image
acquired from USGS had the two scenes mosaicked together. The image was reprojected
into UTM WGS84 37S, to match the coordinate system of the points collected in the
Hield. No further image pre-processing was done to the 2003 image and the Einal result
was a 6 band image made up of three visible bands, one near infrared band and two mid-
infrared bands. The band widths can be found in Table 1. The thermal and panchromatic
bands were not included because I did not think they would add enough useful
information. The 2003 image was used as the reference scene for geometrically and
radiometrically correcting the 1986 and 1995 images.
Table 1. Bandwidths for Landsat 7 ETM+ and Landsat 5 TM satellite imagery
Band ETM+ TM
1 (blue) 0.45-0.52 0.45-0.52
2 (green) 0.53-0.61 0.52-0.60
3 (red) 0.63-0.69 0.63-0.69
4 (NIR) 0.78-0.90 0.76-0.90
5 (MIR1) 1.55-1.75 1.55-1.75
7 (MIR2) 2.09-2.35 2.08-2.35
1986 and 1995: The 1986 and 1995 images are Thematic Mapper images from
Landsat 5 downloaded from the University of Maryland Global Land Cover Facility. The
two scenes that make up the study site were downloaded separately and mosaicked
together using a feathering process to blend the areas of overlap. The images were
geometrically corrected to the 2003 image using 50 to 60 points with a final RMS value
of less than 0.25 pixel and reproj ected to the same coordinate system as the 2003 scene.
The histograms for all six bands of the 1986 and 1995 images were matched to the 2003
image and radiometric corrections were performed on both images using the method
described in Jensen (1996) as multiple-date empirical radiometric normalization using
regression to reduce differences between them and the 2003 image caused by
atmospheric attenuation. Nineteen radiometric control points were chosen so that they
fell on areas that did not change spectrally over time, generally permanent lakes, patches
of bare soil, rock and roads. Digital numbers were recorded for bands 1-5 and 7 and a
linear regression analysis performed on each. R2 ValUeS were all greater than 0.9. Final
equations and R2 ValUeS are given in Table 2. The resulting equations were applied as the
correction to the image.
Both images were subset to an area slightly greater than the boundaries of the
Nakuru Wildlife Conservancy. I was unable to follow the exact boundaries of the NWC
because I was not able to acquire information that would allow me to Eind them
accurately enough on the image.
Table 2. Regression equations used to radiometrically correct the 1986 image to the 2003
Band Correction equation R2 ValUe
1 y=0.943751x-30.2539 0.9240
2 y=1.529852x-16.0622 0.9298
3 y=1.212533x-11.2657 0.9061
4 y=1.030524x+2.8338 0.9506
5 y=0.582207x+5.1756 0.9685
7 y=0.778132x+6.0936 0.9400
All three scenes were classified separately using a combination of the supervised
and unsupervised classification methods (Jensen, 1996). In principle, each landcover has
a unique spectral reflectance in the bands that make up the image. These differences in
spectral reflectance can be used to classify an image by selecting and grouping together
those pixels with similar signatures. When doing a supervised classification, the user
creates training signatures by defining training sites on the image which delineate known,
homogeneous ground covers. Training sites are selected based on ground truthing data,
shapes associated with specific ground covers and in situ knowledge of the area. Ground
truthing data were collected in the Hield by recording the coordinates of landcover patches
with a minimum size of 100 m x 100 m using a Garmin 76 GPS receiver. Other
information collected included dominant vegetation type, ground cover type, open or
closed canopy, and landuse where applicable. When a coordinate could not be taken
from the center of the patch, distance and direction to the center of the patch were
In an unsupervised classification, the computer program divides the pixels into the
specified number of classes based on spectral similarities without reference to outside
sources of information. The steps outlined below describe the classification process for
each of the three images. There were slight differences in the number of steps necessary
for complete image classification between years.
* A normalized difference vegetation index (NDVI) was created from bands 3 and 4
to show differences in vegetation biomass. This calculation (B4-B3/B4+B3) has
been shown to be useful in distinguishing land cover types with varying amounts of
vegetation (Jenson, 1996).
* A tasseled cap analysis (TCA) was created to bring out land cover differences in
brightness (layer one), greenness (layer two) and wetness (layer three). I discarded
the other layers created by a TCA because they do not yield enough useful
* The NDVI and the first three layers of the TCA were stacked onto the original six-
layer image resulting in a 10-layer image used in all subsequent analyses.
* A principal components analysis (PCA) was performed on the 10-layer image. The
maj or components of the first principal component were band 5, the NDVI and the
brightness layer of the TCA and accounted for over 95% of the variation in the
images. A 3-class ISODATA unsupervised classification was performed on the
PCA, set to 20 iterations and a convergence value of 95%. The resulting classes
included water; areas of heavy, healthy vegetation (forest); and areas of low green
biomass and high soil or senescent vegetation (savanna). The three classes were
used to create masks to break the 10-layer image into two images to be further
classified separately. Forest was left on its own while the water and savanna
classes made up the second section.
* A 15 class ISODATA unsupervised classification was performed on the forest
section following the parameters outlined above. Each of the resulting classes was
inspected and placed into one of three categories: badland, bush or forest, based on
their spectral signatures and knowledge of the area. Each category was recorded to
1, 2 or 3 and used to create masks to further break the forest image into 3 images to
be further classified separately.
* Similar steps were followed for the savanna section of the original image except
that a 20 class unsupervised classification was performed and the resulting classes
placed into one of six categories: water, urban, bush, grass, baresoil and mud.
* At this point, some of the individual pieces were left as they were because they
represented only one landcover type. The remaining pieces were either further
subdivided by the method already outlined, or classified using a supervised
classification scheme. Whether a supervised or unsupervised classification scheme
was applied next depended on how many landcover types were represented in the
* For the supervised classifications, training sites were taken from ground truthing
data and familiarity with the area. They were created by outlining the edges of the
training area or by using the area of interest (aoi) seed tool set to a spectral
euclidian distance of 10. The aoi seed tool is used to collect neighboring pixels
with similar spectral characteristics to create a training area. When the seed pixel is
chosen, the program inspects each neighboring pixel to see if it is spectrally similar
based on parameters set by the user. If a pixel is similar, the program then inspects
the neighbors next to it. This continues in a stepwise process until all similar,
neighboring pixels are selected or until the maximum number of pixels is reached.
* Once each piece was classified, they were recorded to one of 9 landcover types:
developed, agriculture, bush, badland, open forest, closed forest, baresoil, grassland
and water. Descriptions of types are given below.
* Finally, the pieces were mosaicked together using the maximum overlay function
to produce the final, complete classified image.
A second set of classified images was created by collapsing the nine classes already
created into three based on the ability of cheetahs to exploit them. These classes are as
follows: suitable (grassland), marginal (bush, open forest and baresoil) and unsuitable
(badland, closed forest, developed, agriculture and water). Further descriptions are given
The classified images were further processed to remove isolated pixels that were
most likely misclassified and within a matrix of dissimilar pixels. An accuracy
assessment of the 2003 image showed that this process did not improve overall accuracy
but rather made it worse by about 3%. However, the 1986 and 1995 images had greater
problems with isolated pixels and I believed that while removing them did not improve
the 2003 image, it would improve the other classification. For this reason, I decided to
reassign the class of all clumps less than 3 pixels in size to the value of the maj ority of
surrounding pixels. The 2003 image needed to be processed in the same way to make
comparisons between years possible.
The accuracy of the 2003 image was assessed by randomly selecting 256 points and
determining their actual landcover based on field work and knowledge of the area. This
was then compared to the classified image using the accuracy assessment function in
Erdas Imagine 8.6.
The resulting four classified images were each analyzed separately using
FRAGSTATS 3.3 to examine differences in classes between and within years. The
following indices were used: (1) total area (ha); (2) number of patches; and (3) mean
patch size (ha). Edge metrics were not considered because cheetahs are landscape
species and are not confined to a single habitat type. They readily move between
landcovers and exploit multiple habitat types.
Representative signature histograms for eight of the nine landcover classes are
shown in Figure 2. Agriculture was not included. Each crop will have a unique signature
so graphing them would be difficult and uninformative. Figure 3 shows pictures of the
Closed forest: This landcover type is dominated by yellow fever acacia (YFA),
Acacia xanthophloea, in the NWC area. It grows to 25 m in height and is most
commonly found around lakes, rivers and in areas with high ground water and black
cotton soil. YFA generally grows in single species stands. Other tree types that may be
found in this landcover category include blue gum (Eucalyptus globuhus), pine (Pinus
sp.), euphorbia (Euphorbia bussei, E. candelabrum and other sp.), and deciduous mixed
hardwoods, though YFA are by far the most common. The hardwoods are confined to
hills and other areas that have not yet been exploited for agricultural or settlement
purposes. Closed forests have an understory that is difficult to penetrate and which
hinders movement. It is dominated by the shrub Phechea bequaertii, a symbiotic species.
Open forest: Open forests are made up almost exclusively of YFA. It differs from
the closed forest category in that the understory is dominated by grasses, forming an
open, parkland landscape. Like the closed forest category, open forest is usually found
near lakes, rivers and other waterways.
SpectralProfiles of Landcover Classes
I Il \ // / I Open Forest
8 Closed Forest
.Z 10 I Developed
50 / I Water
1 2 3 4 5 6 7 8 9 10
Figure 2. Spectral profiles of eight landcover classes
Grassland: Grassland areas are dominated by grass species with some low-lying
herbaceous plants present. Much of this landcover type consists of dead biomass and
patches of bare soil though green grass can be found around Lake Naivasha.
Bushland: Bushland areas are dominated by shrubs, bushes and low-lying trees
interspersed with grasses and patches of bare soil. Bushland and grassland areas are
continuous with varying proportions of bush to grass. The most common type of
bushland found in the conservancy is mixed stands of leleshwa and acacia sp. with
proportions varying in different areas. Other types of bushland include croton, grewia
Agriculture: Agriculture is areas of landscape modified for the purpose of
growing crops, either at the subsistence level or for commercial purposes. Agriculture
areas may also contain some structures such as single-family homes, storage sheds or
buildings used to support the industry.
Developed: Developed areas are landscapes consisting of extensive human
modifications dominated by built structures where the landscape is unrecognizable from
its original form. Examples include urban areas, villages, flower farms (greenhouses),
roads, buildings and other similar structures.
Water: Lakes, rivers and streams.
Badlands: Badlands are areas of thick bush found specifically on old lava flows
and other types of rocky outcrops. There are many species of bushes and herbaceous
species that occur in this landcover type including Aloe sp., Croton sp., Euphorbia sp.,
Acacia sp., Grewia sp., Rhus sp. and others. Grasses in this type of landcover are sparse.
Areas of badland are very difficult to penetrate due to the thickness of the bush and the
spines and thorns associated with them.
Bare soil: Areas of bare soil include mudflats found around the lakes, degraded
lands and cleared patches around urban areas.
The landcover types already described were further refined to give three land-use
types relevant to cheetahs. The landcovers were placed in one of three categories:
suitable, marginal or unsuitable.
Suitable: This category includes those areas where cheetahs are most likely to set
up residence for an extended period of time. Cheetahs are creatures of the open plain.
They are more particular about the areas they choose to live than other large African
felids and require habitats that meet their unique needs. Two features high on the list of
priorities are open space to see their prey and the room to run to attain the high speeds
necessary for capturing prey. Grassland is the only classification that meets these two
requirements. They are also the areas where game is most abundant. Grassland is the
only landcover type that would be considered suitable for cheetahs.
Marginal: Marginal areas are those where cheetahs are known to be found but
they support a lower density of preferred prey species, making them unable to support as
large a cheetah population as found in grassland areas. Marginal areas can also be used
by cheetahs as movement corridors. Land cover types that make up this category include
bushland, badland, bare soil and open forest.
Unsuitable: These landcover types would be avoided by cheetahs and include
developed, closed forest and agricultural areas. Cheetahs are shy creatures and avoid
areas with high human presence as found in agricultural and developed areas. Closed
forests and badlands would also be avoided as they would be too thick for a cheetah to
easily pass through.
Prey Density Estimates
Information on the density of prey within different habitat types was collected
using line transects. It was necessary to determine the density of potential prey species in
the NWC to see if a decline in prey availability could account for the loss of cheetahs and
to determine suitability of different land cover classes for supporting cheetahs. For this
__ r A
Figure 3. Representative images of seven landcover classes. A) Grassland. B) Bushland.
C) Open forest. D) Closed forest. E) Bare soil and water. F) Developed. All
pictures taken by M. Evans
Figure 3. Continued
analysis, I began by conducting a series of transects using the protocols as described by
Buckland et al. (2001) for use in the program DISTANCE 4. 1 (Thomas et al., 2003), a
computer program that analyzes transect data to determine animal densities. For
DISTANCE to give accurate results, three fundamental assumptions must be met when
designing and conducting the survey: (1) "Objects on the .. line are always detected";
(2) "Obj ects are detected at their initial location, prior to any movement in response to
the observer"; (3) "Distances (and angles where relevant) are measured accurately .. or
obj ects are correctly counted in the proper distance interval" (Buckland et al., 2001; pg
I began by randomly choosing one thousand sets of geographic coordinates within
and immediately surrounding the study area using the program ArcView 3.2 to meet
another assumption of DISTANCE, that transects are laid out randomly. The points
served the dual purpose of use as testing sites for the satellite image classification. I
chose the large number because I knew not all points would be accessible or usable. I
also did not have exact coordinates for the study site and knew some of the points would
fall outside of the NWC. One thousand compass points were then randomly chosen by
the program Microsoft Excel and paired with the geographic coordinates. Three habitat
types were censused with transects including grass, bush and open yellow fever acacia
forests. These habitat types were most likely to be used by cheetahs based on their
vegetative structure and potential prey populations. Other habitat types available in the
area were either too developed or too thick to be included in the census. I decided that
the likelihood of a cheetah using a developed or agricultural area regularly was low
enough that I could safely classify those areas as unsuitable without doing prey transects
in them. It was also very clear that developed areas had high human densities and
generally low wildlife densities. Natural areas that were not censused included badlands
and forests with thick undergrowth where it would be impossible to ensure that
assumptions one and two were not violated, and where safety from buffalo was a
concern. These are also areas cheetahs are not known to inhabitat.
Transects were conducted in suitably large patches of individual habitat that could
possibly be used by cheetahs, with the starting point determined by the random point
closest to an edge of the habitat. Walking direction was determined by the compass
direction associated with the starting point. On at least one occasion, the perpendicular
direction was used due to ease of walking. At least two observers were used at all times.
When an animal of interest was sighted, information on species, number (cluster size),
sex, age, UTM position of observer, distance from observer transectt line), and angle
from north was recorded. Distance and angle were recorded to where the animals were
first sited. Angles from north were later converted to angle from the line. Transects were
walked at a speed of 1-2 km per hour from one end of the habitat type to the other, at
which point a parallel line was walked in the opposite direction, offset from the first by
200-500 meters, depending on line of sight distance.
This process was repeated until the entire habitat area was covered, or until it
became too dark to continue. Care was taken to avoid double counting animals on
different transect legs. It was noted whether or not groups moved in the direction of the
next transect leg. If so, then any group with a similar size and composition found on the
next leg was not counted. This was rarely an issue.
Thirteen transects were walked for a total of 19.4 km in grassland habitats and 35.9
km in bush/ open forest habitats. Transects were walked in the early morning hours
starting at about 6:30 hour or in the afternoon starting at about 16:30 hour. These are the
times when wildlife is most active and therefore most likely to be seen. It is also when
cheetahs most actively hunt. Transect data were analyzed using the program
The species censused were those found in the area and known to be preyed upon by
cheetahs. They included zebra (Equus burchelli), eland (Tragelaphus oryx), waterbuck
(Kobus ellipsiprynanus), kongoni (Alcelaphus buselaphus), thompson's gazelle (Galzella
thonasoni), grant's gazelle (Galzella granti), impalas (Aepyceros nzel~anus), warthogs
(Phacochoerus aethiopicus), hares (Lepus capensis), guinea fowl (Nuntida nzeleagris),
steenbok (Raphicerus canspestris) and dik dik (Madoqua kirkii) (Graham, 1966;
McLaughlin, 1970; Eaton, 1974; Burney, 1980, Frame, 1986; Caro, 1994). Thompson's
gazelle, grant's gazelle and impalas are considered primary prey species for this study as
they represent between 62% and 75% of cheetahs' diet in studies conducted in Kenya
(McLaughlin, 1970; Eaton, 1974; Burney, 1980).
Data required for DISTANCE to do the analyses include transect number, cluster
size, radial angle from line (this was calculated based on angle of the line from north),
distance from line and total length of transect. Total length of transect was calculated by
adding up the lengths of the individual legs of each transect. All analyses were based on
cluster size as most of the species recorded occurred in groups rather than as individuals.
Counts in open YFA and bush were pooled together as there were not enough sightings in
either habitat type to give good results. Also, the habitats had similar characteristics.
The sightings were divided into different groups based on body size and habitat. Groups
analyzed included: large herbivores (>40 kg) in grass, large herbivores in bush,
thompson' s gazelle in grass and bush, impala and grant's gazelle in grass and bush, small
herbivores (<12 kg) in grass and bush and preferred prey in grass and bush. Large
herbivores included zebra, eland, waterbuck and hartebeest. Small herbivores included
hares, dik dik, steinbok and guinea fowl. Warthogs were censused but not included in the
analyses as they did not fit easily into any of the categories and there were not enough of
them to analyze separately. DISTANCE assumes that the animals being counted have the
same probability of detection if they are being analyzed together. I believe that this
assumption is not violated with the groupings I have made as the species in each group
are "of similar size and provide similar visual cues" (Buckland et al., 2001: pg 302).
For the analyses, a natural log transformation was applied in DISTANCE to the
cluster size for all groupings because of the large variation seen in cluster size between
groups. The transformation reduces the influence of a few large cluster sizes on the
estimation of density. For those analyses done on groupings in the bush, perpendicular
distance from the line, yi, was replaced with g(yi) in the regression. "G(yi) is the
estimated detection function from the fit of the selected model to the distances from the
line or point to detected clusters." (Buckland et al., 2001). This corrects for the problem
of a shoulder in the detection function where "mean cluster size does not increase with
distance until detection distance exceeds the width of the shoulder." (Buckland et al.,
2001). The same correction was not applied to the groupings in grass habitats because
detection did not vary with distance due to the nature of the terrain. Only the small bush
group was truncated 5% at the right tail due to some distant outliers that made modelling
the regression line problematic.
Five models were run for each of the groupings and Akaike's information criterion
(AIC) values compared to choose which model gave the most robust results. In those
instances where AIC values differed by less than two, model choice was also based on
goodness of fit tests provided by DISTANCE. Those models in each of the groups with
the lowest AIC values and the best fit were chosen to report results (Table 3) (Buckland
et al., 2001).
Table 3. DISTANCE models used in analyses
Key function Series expansion
Uniform Simple polynomial
Half-normal Hermite polynomial
Hazard-rate Simple polynomial
Differences in densities between bush and grassland habitats for each group were
tested using a T-test as described in Buckland et al. (2001). The biomass of preferred
prey species in each habitat type was calculated from the density. The total biomass for
the group was calculated by multiplying the density estimate for the group by the average
body mass of the group. Average body mass was calculated by multiplying the number
of times a species was seen in a habitat by that specie' s average body mass as reported by
Schaller (1972) and Caro (1994), adding up the body masses for all individuals and
dividing the total by the total number of animals from that group.
Potential Cheetah Population Size
The potential number of cheetahs the NWC could support was calculated using two
methods. I used the regression equation of Gros et al. (1996) to predict cheetah biomass
per unit area based on prey biomass (y = 0.002x + 0.21) in the Nakuru area. I also
followed Emmons (1987) by assuming that 70% of a prey individual was edible and that
cheetahs could take 10% of the prey population per year. I divided the resulting available
biomass by a cheetah's average yearly food requirement (calculated from the average
daily requirement of 4 kg/day (Schaller, 1972)) to again determine the potential cheetah
population size for the NWC.
NWF and LNNP Counts
The Nakuru Wildlife Forum has conducted biannual game counts since October of
1996 throughout the conservancy area, with the exception of 2001. Counts on all
properties are conducted on the same day at the same time to reduce the possibility of
double counting animals. The different properties are divided into blocks based on
configuration of roads, vegetation and area to be covered. Teams of counters consisting
of at least one driver, one spotter and one recorder drive through each block and get total
counts of all mammal species encountered. Except for some Lake Nakuru National Park
(LNNP) counts, information about sex or age of animal, and vegetation type is not
recorded. The block counts are later compiled into total counts for each property and for
the total conservancy to track changes in game counts over time. Only the fall counts
were used for this study. There was some evidence of biannual fluctuations in animal
numbers so only those counts that coincided with the timing of the density counts were
used. Because the number of properties participating in the counts, and thus the total area
surveyed, varied over time, total counts of wildlife numbers could not be used for
comparison purposes. Instead, yearly average density was calculated by dividing the
total number of animals counted throughout the survey by the total area surveyed. In
order to detect statistically significant changes in density, average density was regressed
against year using Microsoft Excel. Individual species were tested as well as preferred
prey as a group and large prey as a group. Small prey were not tested. Not all of the
species included in the DISTANCE analyses were counted as part of the forum counts.
Also, forum counts are usually conducted from a vehicle rather than on foot, reducing the
likelihood of accurately counting small mammals.
The Einal classifications for the 1986 and 2003 scenes can be seen in Figures 4
through 7. The 1995 classification was not included because the results of the
classification did not make sense, which I think is related to problems with the
classification itself rather than actual landcover changes. The overall accuracy for the
nine class 2003 classification was 70.3%. Kappa statistics and producers and users
accuracy are summarized in Tables 4 and 5. The errors associated with the classification
occurred between grass and bush, grass and agriculture, closed forest and open forest,
urban and baresoil, and bush and open forest. Forty of 76 misclassified points occurred
between grass and bush. These are continuous habitat types and defining where the
cutoff between grass and bush occurs spectrally was difficult. I'm sure that those areas of
pure bush or pure grass were correctly classified, but areas with sparse bush or thick grass
are the areas most likely to have been classified incorrectly. Six of the misclassifications
occurred between grass and agriculture. A common crop grown in the NWC is wheat,
making grass and agriculture easy to separate when creating training sites based on shape
of agricultural plots, but difficult to separate spectrally. Their signatures were quite
similar and therefore often confused in the final classification. Bare soil in the NWC is
often very lightly colored, and local soils can be used as building materials, making the
spectral signatures for these classes similar. It was impossible to sufficiently separate
developed from baresoil areas despite repeated attempts, especially along lake shores.
fit" .; CI
,, r Pr
: ~ J
ir r. ~''`
4 j~: d
;I_ ." -I
t `"~ :'4~0~;
Figure 4. 2003 landcover classification. Enhanced Thematic Mapper image ofNakuru
Wildlife Conservancy, Kenya, 04 February 2003.
F' I. 'ir,
.i :. .
G rass lan d
Figure 5. 1986 landcover classification. Thematic Mapper image ofNakuru Wildlife
Conservancy, Kenya, 28 January 1986.
Figure 6. 2003 suitability classification. Enhanced Thematic Mapper image ofNakuru
Wildlife Conservancy, Kenya, 04 February 2003.
0 5 10 20 Kilometers
Figure 7. 1986 suitability classification. Thematic Mapper image ofNakuru Wildlife
Conservancy, Kenya, 28 January, 1986.
The problem with closed forest vs. open forest is similar to grass vs. bush. These
landcover classes occur along a continuum. Also, if open forests have a closed canopy,
they could easily be confused with closed forest. Eleven of the misclassifieations
occurred between bush and open forest. Confusion between these landcover classes is
not surprising considering the similar make up of green trees/ bushes interspersed with
The overall accuracy for the three-class 2003 classification was 73.8%. The
problems with this classification were similar to those for the nine class classification,
most notably the misclassifieation of bush and grass resulting in the misclassifieation of
the suitable and marginal habitats.
Table 4. Accuracy results for 2003 landcover classification
Overall accuracy: 70.3%
Kappa Statistic: .5824
Class Kappa Producers Error Users Error
1 0.6640 100.00 66.67
2 0.5803 25.00 60.00
3 0.4967 74.51 69.72
4 0.8951 75.00 90.00
5 0.8437 27.27 85.71
6 0.2411 33.33 25.00
7 0.8280 62.50 83.33
8 0.4800 77.63 63.44
9 1.0000 100.00 100.00
Table 5. Accuracy results for 2003 suitability classification
Overall Accuracy: 73.8%
Kappa Statistic: .5766
Class Kappa Producers Error Users Error
1 0.4611 77.63 62.11
2 0.5871 72.73 80.00
3 0.7899 70.83 82.93
Habitat suitable for cheetah and its prey decreased between 1986 and 2003 from
1 13,970 ha (46.5% of the landscape) to 95,969 ha (39. 1% of the landscape), respectively.
This is a loss of 18,001 ha, or 15.8% of suitable landscape in a 17-year period. During
this same time, marginal lands increased from 94,891 ha (38.7%) to 108,843 ha (44.4%),
an increase of 13,952 ha, or 14.7%. Unsuitable area also increased, from a low of 36,202
ha (14.8%) in 1986 to 40,339 ha (16.5%) in 2003, an increase of 4,137 ha, or 1 1.4%. The
increase in marginal areas was due to a growth of the bush class and a large increase in
the bare soil class that comprised the marginal category. Bush increased by 10,654 ha
(12.9% increase) while bare soil increased by 6,934 ha, an increase of 290%. The
increase in unsuitable areas was due mostly to an increase in developed areas, from 562
ha to 2,517 ha, an increase of 349%. Table 6 summarizes landcover changes between
1986 and 2003.
There were also changes in patch dynamics over the 17-year period. The average
size of grassland and suitable patches, calculated as the sum of the area of all patches of a
particular type divided by the total number of patches of that type, decreased between
1986 and 2003. Grassland patch size decreased from 16.04 ha to 13.8 ha. The number of
patches for this class remained about the same between years (1986: 7,158; 2003: 7,084).
Suitable patches decreased from 15.4 ha to 12.4 ha. The number of patches also
remained about the same for this category (1986: 7,391; 2003: 7,744).
The marginal category saw an increase in the number of patches, from 7,958 to
9,360, but the average size of the patches remained about the same (1986: 1 1.92 ha;
2003: 1 1.62 ha). Within the marginal category, the number of patches of bush decreased
(9,3 54 to 7,865) but the average size of those patches increased (8.85 ha to 1 1.88 ha),
Table 7. Class metrics for the 1986 and 2003 classifications
1986 2003 1986 2003
Number of Number of Average patch Average patch
Landcover class patches patches size (ha) size (ha)
Suitable 7391 7744 15.42 12.39
Grassland 7158 7084 16.04 13.80
Marginal 7958 9360 11.92 11.63
Bush 9354 7865 8.85 11.88
Baresoil 939 3736 2.55 2.50
Open forest 2954 4963 3.03 1.09
Unsuitable 6758 10440 5.36 3.86
Developed 934 1380 0.60 1.82
Agriculture 3188 4066 1.37 0.97
Badland 2551 4892 3.16 1.88
Closed forest 1107 2582 3.22 2.00
Water 35 101 558.60 181.94
possibly indicating the consolidation and growth of patches originally found in 1986.
Baresoil patches stayed about the same size (1986: 2.55 ha; 2003: 2.50 ha) but there was
a dramatic increase in the number of them, from 939 to 3,736.
For the unsuitable category, average patch size decreased from 5.36 ha to 3.86 ha,
but the number of patches increased from 6,758 to 10,440. Urban areas increased in
patch numbers (1986: 934; 2003: 1,380) and in average patch size (1986: 0.60 ha; 2003:
1.82 ha) (Table 7).
Table 6. Landcover change in the Nakuru
Landcover class 1986 (ha)
Open forest 8954
Closed forest 3569
Wildlife Conservancy, (NWC),
2003 (ha) Change (ha)
1986 to 2003
The density of preferred prey species in grassland habitats was estimated to be
0.628 animals per hectare (95% CI = [0.422, 0.934], CV = 18.7%). In bushland habitat
the density estimate was 0.147 animals per hectare (95% CI = [0.048, 0.454], CV =
57.3%). The preferred prey density estimates are significantly higher in the grassland
habitat than in the bushland areas (T = 3.328, P<0.003). Prey densities for individual
species that make up the preferred prey group are also significantly different between the
two habitat types. Thompson's gazelles are found at a density of 0.442 individual per
hectare (95% CI = [0.273, 0.715], CV = 22.0%) in grass and 0.068 individuals per
hectare in bush (95% CI = [0.022, .207], CV = 57.4%) (T = 3.532, P =0 .003). Impala
and grant's gazelle were analyzed together. Their density in grass was 0.190 individuals
per hectare (95% CI = [0.096, 0.374], CV = 33.9%) and 0.042 individuals per hectare in
bush (95% CI = [0.010, 0.174], CV = 73.0%) (T = 2.080. P<0.05). The large prey and
small prey groups did not have statistically significant differences in densities between
grass and bush. A summary of density estimates for all groups can be found in Table 8.
Table 8. Density estimates for prey species in grassland and bushland habitats.
Density CV Density CV
Species/ Group (ind./ha) CI (%) (ind./ha) CI (%)
Preferred prey* 0.628 0.422, 0.934 18.7 0.147 0.048, 0.454 57.3
Thompson's gazelle* 0.442 0.273, 0.715 22.0 0.068 0.022, 0.207 57.4
Grant's gazelle and 0.190 0.096, 0.374 33.9 0.042 0.010, 0.174 73.0
Large prey 0.138 0.070, 0.271 34.4 0.097 0.045, 0.207 36.5
Small prey 0.147 0.035, 0.613 68.3 0.342 0. 180, 0.649 31.1
* indicates statistically significant differences between grassland and bushland densities.
The large coefficient of variation seen in some of the density estimates is due to a
low number of sightings of individuals in those groups. It is also due to the patchy
distribution of individuals in the habitats. Some species were counted on some transects
but not on others. For example, thompson's gazelles were counted on all transects for the
grassland areas but were missing from three transects conducted in bush. This pattern
was also found for large prey. Grant' s gazelle and impala were missing from five bush
transects but were counted on all grass transects. DISTANCE gives unbiased results
even with low sighting numbers so the results are valid. Though significant differences
were not found in the group densities of large prey and small prey between habitat types,
this could be due to the large CV. More transects may have reduced the coefficient of
variation and pulled out differences in these groups, especially the small prey groups.
However, the small prey that make up the group have an average body mass of less than
4 kg. Even if there were significant differences in densities, it is unlikely that the
differences in biomass between habitats, when combined with the other groups, would be
enough to significantly impact a cheetah's decision to exploit grassland versus bushland
The biomass of preferred prey in suitable areas is 1505 kg/km2 (95% CI = [101 1,
2238]). The biomass of preferred prey in marginal areas is 384 kg/km2 (95% CI = [125,
1186]). Using Gros et al. (1996), in 1986, the NWC could support a potential cheetah
population of 73 (51-107) individuals in suitable areas and 19 (9-49) in marginal areas for
a total of 92 (60-156) cheetahs. In 2003, suitable areas could support 62 (43-90)
individuals while marginal areas could support 21 (10-56) individuals for a total
population of 83 (53-146) cheetahs.
Using the method espoused by Emmons (1987), suitable areas could support 85
(57-127) cheetahs while marginal could support 18 (6-56) in 1986 for a population total
of 103 (63-183). In 2003, the potential cheetah population size in suitable areas was 72
(48-107) individuals and 21 (7-64) in marginal areas for a total of 93 (55-171) cheetahs
Table 9. Potential cheetah population estimates as predicted by prey biomass.
Method Year Habitat
Suitable (range) Marginal (range) Total (range)
Gros et al. 1986 73 (51-107) 19 (9-49) 92 (60-156)
(1996) 2003 62 (43-90) 21 (10-56) 83 (53-146)
Emmons 1986 85 (57-127) 18 (6-56) 103 (63-183)
(1987) 2003 72 (48-107) 21 (7-64) 93 (55-171)
Prey Density Changes Over Time
In no instances did population densities increase between 1996 and 2003. Rather,
all populations exhibited a decrease of some degree, though regression analyses of
changes in prey density indicate a significant decline in only two instances. There has
been a significant decline in the density of preferred prey since 1996 from a density of
0.2097 animals per hectare to 0.1207 per hectare (R2 = 0.661, slope = -0.0171, SE =
0.0061). This is due mostly to a decline in thompson' s gazelle from 0. 1186 animals per
hectare to 0.0678 per hectare (R2 = 0.749, slope = -0.0088, SE = 0.0025). Trends for
other species were not significant at the 0.05 level. Three trends: large prey, impala and
kongoni, were significant at the 0.1 level (large prey: R2 = 0.623, slope = -0.0185, SE =
0.0072; impala: R2 = 0.592, slope = -0.0051, SE = 0.0021; kongoni: R2 = 0.638, slope =
-0.0052, SE = 0.0020). A summary of the results for all analyses are given in Table 10.
It should be noted that grant's gazelle are often mistaken for thompson' s gazelle, so the
absolute numbers reported in the census may over-represent thompson' s gazelle and
under-represent grant' s gazelle. However, thompson's gazelle are more common, and I
doubt that misidentification occurs often enough to significantly change the results. As
more counts are conducted in the NWC area, incidences of significant declines in species
may increase. Also, the results from the forum counts should be viewed with some
caution. They are designed to count all animals within the census area, but the accuracy
of the method is not known, nor is the detection probability for the different species being
censused. There is no way of knowing what proportion of individuals of each species is
missed in the counts. Also, the participants and their level of experience vary from year
to year, and there is no way to evaluate inter-observer differences either within or
between years. However, the same methods are applied for every count, therefore the
results should be comparable and trends over time should be indicative of overall
Table 10. NWC wide prey density estimates and regression analyses of density changes.
(ind./ha) R2 Slope SE
1996 1997 1998 1999 2000 2002
Preferred prey** 0.2097 0.2261 0.1398 0.1272 0.1398 0.1207 0.6610 -0.0171 0.0061
Thompson's gazelle** 0.1186 0. 1224 0.0819 0.0830 0.0867 0.0678 0.7492 -0.0088 0.0025
Impala* 0.0715 0.0655 0.0479 0.0353 0.0427 0.0434 0.5918 -0.0051 0.0021
Grant's gazelle 0.0197 0.0382 0.0100 0.0089 0.0104 0.0095 0.3554 -0.0032 0.0021
Large prey* 0.1785 0.1983 0.1054 0.0799 0. 1004 0.0854 0.6227 -0.0185 0.0072
Zebra* 0.1189 0.1223 0.0697 0.0598 0.0714 0.0613 0.6348 -0.0106 0.0040
Eland 0.0187 0.0296 0.0158 0.0073 0.0132 0.0109 0.4060 -0.0023 0.0014
Kongoni* 0.0313 0.0388 0.0129 0.0057 0.0094 0.0061 0.6376 -0.0052 0.0020
Waterbuck 0.0096 0.0076 0.0070 0.0071 0.0064 0.0071 0.5118 -0.0004 0.0002
**indicates declines which are statistically significant at the 0.05 level. *indicates declines which are statistically significant at the
Cheetah Population Trends
In 1975, Myers (1975b) estimated the population of cheetahs to be about 15,000
animals throughout their range, possibly half the population size from the 1960's. The
Kenya population at that time was estimated to be less than 2,000 animals and under
pressure from loss of habitat due to exploitation of rangelands for agriculture and
livestock ranching. Gros (1998) estimated, based on available habitat and prey densities,
the potential Kenya cheetah population at 10,000, but the actual number was probably
closer to 1,000 to 2,000 animals. According to Gros (1998), the Kenya population
overall has most likely remained fairly stable. However, in the Nakuru area, the cheetah
population appears to have declined. In 1990, Gros (1998) estimated the population to be
around 3 5 animals based on interviews, with the maj ority of the respondents reporting a
decrease. Using the same interview technique in 2002, the Cheetah Conservation Fund
reported the Nakuru cheetah population to be about 12 animals (Wykstra, pers. comm.).
The interview technique has been shown to be the most reliable indirect method for
estimating densities of large carnivores (Gros et al., 1996).
Using the averaging technique (Gros et al., 1996), Gros estimated the cheetah
population within Lake Nakuru National Park to be about three animals. In 1996, KWS
counted two cheetahs within the park during one of their triannual censuses. One cheetah
was counted in 1997. Since then, no confirmed cheetah sightings have been reported and
the LNNP cheetah population is believed to be lost. It should be noted that cheetahs do
not persist in areas with high densities of other carnivores, especially lions. Lions kill
cheetahs and their cubs and are responsible for 73% of the mortality of cheetah cubs in
the Serengeti (Laurenson, 1994). Lions were translocated into LNNP during the 1990's
and their population flourished. Twenty-seven lions were seen during one of the 2000
game counts. Game numbers remain high within the park so the loss of cheetahs is most
likely attributed to the lion population rather than a loss of prey species. But the LNNP
cheetah population could recover. In 2002, two rangers were killed by lions within the
park and as a result the lions were killed, only lionesses with cubs were allowed to
remain per Kenya law. It is thought that the remaining females will also be killed once
the cubs are grown. With the loss of the lions from the park, cheetah recovery in that
area is possible. While no confirmed sightings of cheetah within the park have been
reported since 1997, one questionable cheetah sighting was made in the fall of 2002.
There is some controversy about whether it was actually a cheetah or a leopard.
There is also some indication that the distribution of cheetahs within the Nakuru-
Naivasha area has changed between 1990 and 2002. Gros (1998) reported 20 cheetahs in
the properties north of Lake Naivasha, and 15 individuals in the properties south of it.
While the CCF report does not give separate abundance estimates for the two areas, it is
clear from the list of sightings that the maj ority, especially frequent or regular sightings,
occurred in the area south of Lake Naivasha. Table 11 summarizes sighting information.
Suswa, Hell's Gate National Park, Kongoni Game Sanctuary and Kedong all report
regular sightings of cheetah on a weekly or monthly basis. Suswa reports regular
sightings of mothers with cubs, most likely more than one family group are present on the
property. In contrast, sightings on properties north of Lake Naivasha generally consist of
only one or two individuals seen only once or twice.
Table 11. Cheetah sightings in the Nakuru Wildlife Conservancy: 2000-2002
Group size and Seen
Area of NWC Property composition regularly?
North of Lake Kekopey Group Ranch One individual No
Naivasha Kigio One individual No
Kigio Two individuals No
Marula Two individuals No*
Mwariki One individual No*
Soysambu Two individuals No
Soysambu One individual No
South of Lake Suswa Mother with 6 cubs Yes
Naivasha Suswa Family groups of 3-4 cubs Yes
Suswa Indivi dual s Yes
Suswa Family groups up to 5 cubs Yes
Kongoni Game Valley One individual No
Hell's Gate NP 2-4 individuals Yes
Hell's Gate NP One individual Yes
Kedong One individual Yes
Kedong Mother with 2 cubs Yes
Kongoni Game Sanctuary One individual No
Kongoni Game Sanctuary Two individuals Yes
* seen twice
Though not quantified, landcover change in the Nakuru Wildlife Conservancy
shows some important trajectories. Grassland in many areas of the conservancy has been
replaced by the less productive (in terms of ungulate densities) bushland and baresoil
categories. The growth of baresoil is especially apparent along the southern portions of
Lake Elementaita, to the east of Lake Nakuru National Park, to the southeast of
Elementaita town and the very southern portion of the study site. While overgrazing and
poor land management practices have been implicated in the degradation of grassland
areas (Milton and Dean, 1995; Kellner and Bosch, 1992), stocking rates of livestock for
individual properties in this livestock area were not recorded. However, they are likely
candidates for the increase of baresoil.
The growth of bush and other woody species in grasslands, termed bush
encroachment (Moleele and Perkins, 1998), is a problem faced by land managers
worldwide. In Botswana (Moleele et al., 2002), bush encroachment has been implicated
in the loss of high quality rangeland while its growth in savanna areas of South Africa
(Roques et al., 2001) has been widely observed. In the Nakuru area, this phenomenon is
most apparent in the areas north of Lake Naivasha and between Lakes Nakuru and
Elementaita. Livestock ranching has been shown to increase the rate of bush
encroachment (Brown and Archer, 1989; Hudak, 1999). The spread of bush in
grasslands occurs most readily where cattle grazing occurs. The Madikwe Game Reserve
in South Africa experienced a 30% relative increase in bush during a 40-year period due
in part to long-term cattle grazing in the area (Hudak and Wessman, 1996). A study of
shrub encroachment in Swaziland showed that grazing pressure was a key determinant in
the spread of woody species in a lowveld savanna (Roques et al., 2001). Bush
encroachment has been shown to radiate out from focal points such as a paddock or water
trough, a trend found on a cattle ranch in Tanzania (Tobler et al., 2003). Bushland cover
decreased from areas of high cattle intensity to the more extensively used game reserve.
Moleele and Perkins (1998) examined fifteen environment variables to explain bush
encroachment in Botswana and found that high cattle density was responsible for bush
encroachment around boreholes and cattle troughs. Bush encroachment in the Nakuru
area is most likely to be caused by cattle due to the importance of this land use in the
Another important land-use change for the area is the growth of developed areas.
Nakuru town has grown extensively along with Naivasha town. But most obvious is the
increase in developed areas along the southern shore of Lake Naivasha. The commercial
flower industry has grown enormously since 1986; almost no trace of it can be found in
the 1986 image. The developed areas seen in the 2003 images are greenhouses and
housing for the thousands of workers who support the industry.
Direct Consequences of Landcover Change
It is unlikely that the loss of grassland habitat or the increase in bush have had any
direct negative effects on the Nakuru cheetah population. While cheetahs prefer open
grasslands, they are able to use a wide variety of habitat types, from open grassland to
heavy bush. In fact both are necessary to provide food and cover from predators and the
heat of the day (Caro and Collins, 1987; Schaller, 1972). In Karamoja region of Uganda,
cheetahs prefer open habitats with less than 50% woody cover and grasses of medium
length (Gros and Rejmanek, 1999). In Nairobi National Park in Kenya and Serengeti
National Park in Tanzania, cheetahs use the grassland areas but are also found in the
woodlands (Eaton, 1974, Caro, 1994; Schaller, 1972). Myers (1975b) and Hamilton
(1986) report that cheetahs are frequently found in bushlands, often because other, more
suitable habitats are not available. It is also unlikely that changes in the configuration of
marginal and suitable habitats have had a negative effect. In areas where the appropriate
habitat makes up more than 20-30% of the total landscape, patch configuration and
arrangement are of less importance than habitat amount (Andren, 1994; Fahrig, 1997).
This is truer for generalist species and landscape species able to move through less
appropriate habitat types to reach suitable patches than for habitat specialists or species
sensitive to spatial-temporal pattern of patches (Sunquist and Sunquist, 2001). In 2003
suitable habitat comprised 3 9% of the landscape while marginal comprised over 44%,
both above the threshold level of 30%. Fragmentation is unlikely to play a role in
determining the Nakuru cheetah population size until the availability of appropriate
habitat, both suitable and marginal combined, falls below that 30% threshold level.
Of greater importance is the increase in the amount of unsuitable habitat, especially
developed areas. Cheetahs are shy (Schaller, 1972; Hamilton, 1986) with low
competitive ability against competitors (Durant, 2000; Durant, 1998). They are less
tolerant of human presence than other carnivores and are therefore more likely to avoid
areas of high human density. The human population in the Nakuru District has increased
by more than 300% between 1969 and 1999. Average density was 137 people/ km2 in
1999, up from 41/ km2 in 1969 and 1 18/ km2 in 1989, with pockets of greater densities
centered around the towns and areas of small scale agriculture. Evidence for the increase
can be seen along the southern edge of Lake Naivasha where the area of development has
increased from almost nothing to running the length of the shore. The area covered by
the three principal towns has also increased dramatically. The influence of developed
areas on cheetahs and wildlife in general extends well beyond the mere conversion of
suitable habitat to unsuitable. The increase in the human population associated with
increased development results in a regular or constant human presence in areas adj acent
to developed areas. Dogs, lights, noise and traffic all reduce the probability that a
cheetah will exploit areas deemed suitable based on landcover classification alone. As
the human population increases, cheetahs are more likely to be disturbed with consequent
negative effects. Amur tigers have been shown to be more likely to abandon kills and
consume less meat after disturbance by humans (Kerley et al., 2002). Bobcats in
southern California showed little tolerance of urban activities based on the percentage of
their home range composed of developed areas (Riley et al., 2003).
Woodroffe (2001) has calculated the critical human density for which there is a
50% chance of a carnivore population going extinct. Based on Hamilton's (1986)
cheetah survey, Woodroffe has estimated the critical human density for cheetahs in
Kenya to be 16.5 people/ km2, far below the current density in Nakuru. In India,
however, the cheetah population did not go extinct until mean human density reached 120
people/ km2 (in 1901) (Woodroffe, 2001). Clearly there is variation in the ability of
cheetahs to adapt to increasing human densities. The variation is most likely due at least
in part to the amount of persecution and harassment that the cheetahs must contend with,
indicating that cheetahs are not as persecuted in the Nakuru area as they are in other parts
of their Kenyan range. Yet some intolerance exists as cheetahs also come into direct
conflict with humans over resource use. While the members of the NWF are tolerant of
cheetahs on their properties and even encourage their presence, other landowners in the
area are not as favorably disposed. Cheetahs are large carnivores known to kill sheep and
goats (Frank, pers. comm.; Wykstra, pers. comm.; Marker et al., 2003). Many smaller
landowners are unable and unwilling to absorb the cost of livestock losses, and will
harass or even kill carnivores to protect their stock (Frank, 1998; Marker et al., 2003).
Poaching for skins is also an issue in the Nakuru area. Two cheetah skins along with
eight leopard skins were confiscated from a poacher in the fall of 2002 by Kenya Wildlife
Indirect Consequences of Landcover Change
While the conversion of grassland to bushland habitats is of minor importance to
cheetahs directly, it potentially has a much greater indirect effect through food
availability and their preferred prey. Thompson' s gazelles, grant' s gazelles and impala
are found at much greater densities in grassland habitats than in bushland areas. The loss
of grass will necessarily result in a reduction of the prey available to cheetahs. This trend
is evident in the significant decline of thompson' s gazelles since 1995, and in the decline
of impalas. Gros et al. (1996) and Laurenson (1995a) have demonstrated that there is a
"strong correlation" between the biomass of cheetahs and the biomass of their preferred
size class of prey, indicating the importance of prey availability to cheetah success.
The amount of prey biomass available to carnivores has been shown to be an
important determinant of the number of carnivores a given area can support (Fuller and
Sievert, 2001). A positive correlation between prey biomass and carnivore biomass or
density has been found for many carnivore species including cheetahs (Laurenson, 1995a;
Gros et al., 1996), leopards (Stander et al., 1997), lions (Van Orsdol et al., 1985), tigers
(Karanth et al., 2004) and Ethiopian wolves (Sillero-Zubiri and Gotelli, 1995). The
effects of prey depletion can be manifested in a carnivore population in a number of
different ways. With a decrease in prey resources, some carnivores expand their home
range size in response to the reduced carrying capacity. The Triangle region of the Masai
Mara, where prey biomass is comparatively low, supports a lower density of lions with
larger home ranges than the Sekenani or Musiara regions of the Mara (Ogutu and Dublin,
2002). Ethiopian wolves in areas with low prey densities have larger home ranges and
smaller group sizes than wolves in areas with more prey (Sillero-Zubiri and Gotelli,
1995). When prey depletion is modelled in a tiger population, carrying capacity is
reduced, the population size decreases and extinction risk for the population increases
(Karanth and Stith, 1999). Other effects of prey depletion include suppressed breeding or
increased mortality of cubs (hyenas: Holekamp et al., 1999; lions: Hanby et al., 1995; San
Joaquin kit foxes: White and Ralls, 1993; wolves: Boertje and Stephenson, 1992; and
cheetahs: Caro, 1994) and increased mortality of adults (lynx: Poole, 1994).
In cheetahs, it has been suggested that they expand their home ranges in response to
declining prey availability (Caro, 1994). They have also been shown to expand home
range size when prey density is high but patchily distributed with areas of low density in
between (Caro, 1994). The decrease in density of prey species since 1996 combined with
the patchiness of suitable habitat in the study site would result in the reduced carrying
capacity of the NWC for cheetahs. When prey densities decline, cheetahs must travel
further to locate and acquire sufficient food. Increased energy expenditure to obtain food
may have negative consequences on cheetah reproductive rates. Cheetahs generally have
large litters with short inter-birth intervals compared to other large felids. The energy
requirements to successfully raise a large litter to independence are enormous. When
maternal food intake falls below a threshold level of 1.5kg/day, cub growth has been
shown to decline sharply (Laurenson, 1995b). In the Serengeti, 95% of cheetahs die
before reaching adulthood, with the maj ority of mortality due to predation by other large
carnivores. Only 7% of cub mortality can be attributed to starvation or abandonment
(Laurenson, 1994). However, in the Nakuru area, the large carnivores have been either
extirpated or their population size suppressed through human intervention. Therefore,
cub mortality from depredation has been virtually eliminated, putting greater pressure on
cheetah mothers to acquire enough food to raise their large and still intact litters. In this
situation, it is probable that prey density would be a more important determinant in
survival and reproductive rates of cheetahs than has been previously seen in other
sy stem s.
The calculations used to estimate potential cheetah population size based on prey
biomass assume that all grassland and bushland patches are equal in their ability to
support cheetahs and their prey species. Also, these calculations take into account only
preferred prey species, meaning the number the area could support based on total prey
biomass may actually be higher. However, the assumption of equal patch quality is not
realistic. Not all areas classified as grassland or bushland may be appropriate habitat due
to proximity to human habitation or other disturbances. Prey numbers in areas adjacent
to high human densities or with poor security may be depressed due to poaching of prey
species by people (O'Brien et al., 2003). Snares for ungulates are often found along
fence lines and in other areas of high ungulate traffic.
Considering that in 2003 the Nakuru area could have potentially supported a
population of more than 60 cheetahs based on available prey biomass, but the actual size
of the population was estimated at 12 (Wykstra, unpublished report) suggests that prey
depletion is unlikely to be the primary cause of the current decline in the cheetah
population. However, prey densities could become a bigger factor if conversion of
grassland to bushland and degradation of grasslands continues.
The results of this study indicate that recent (1986-2003) changes in landcover and
prey availability within the Nakuru Wildlife Conservancy are insufficient to explain the
marked decline of cheetahs in the area. While grasslands within the conservancy are
converting to less appropriate landcovers due to bush encroachment and degradation,
there is still sufficient habitat and prey available to support a healthy cheetah population.
However, an increase in human density within the conservancy probably plays a
significant role in discouraging cheetahs from using the area to a greater degree (Janis
and Clark, 2002); this problem will worsen rather than improve with time as the area
continues to grow. Support for this possibility is given by the paucity of cheetahs found
in the northern part of the conservancy. Cheetahs may find it difficult to pass through the
densely settled area around Lake Naivasha. Other issues cheetahs face include more
intensive land conversion and subdivision of larger properties surrounding the
conservancy rather than change within the conservancy itself. Subdivided land used for
subsistence farming will convert land from suitable or marginal to unsuitable and present
a barrier to cheetah movement into the area. This landcover change was identified as a
threat to cheetahs in the Nakuru area, especially to the north and west, by Gros (1998)
during her survey in 1990. Fritz et al. (2003) found that wildlife in the Zambezi Valley
of Zimbabwe were less likely to use sections of the river bordered by agriculture. The
negative effect of agriculture on density and diversity of wildlife using the area was
greatly enhanced once a threshold level was reached (Fritz et al., 2003). It is likely that
the growth of agriculture and subsistence farming in the areas surrounding the NWC has
had a similar effect.
Members of the Nakuru Wildlife Forum who wish to see a return of cheetahs to
their property will have to manage their game populations to maintain a high density of
the cheetahs' preferred prey species. They will also need to ensure that poaching of prey
species and of cheetahs is deterred and that cheetahs are not harassed if they colonize
their property. More importantly though, forum members will need to establish and
maintain connectivity between the source population of cheetahs to the south of the
Nakuru area, particularly the Masai Mara, and the rest of the conservancy. The Nakuru
area has never maintained a high cheetah population. More likely the area was used as a
corridor to pass from the southern part of the country to the central highlands and the
Laikipia Plateau. But the growth of settled and densely populated areas may have
reduced or even closed that corridor.
LIST OF REFERENCES
Andren, H. 1994. Effects of habitat fragmentation on birds and mammals in landscapes
with different proportions of suitable habitat: a review. Oikos. 71:355-366
Boertje, RD, Stephenson, RO. 1992. Effects of ungulate availability on wolf reproductive
potential in Alaska. Canadian Journal of Zoology. 70(12):2441-2443
Brown, JR, Archer, S. 1989. Woody plant invasion of grasslands: establishment of honey
mesquite (Prosopis glan2dulosa var. glan2dulosa) on sites differing in herbaceous
biomass and grazing history. Oecologia. 80:19-26
Buckland, ST, Anderson, DR, Burnham, KP, Laake, JL, Borchers, DL, Thomas, L. 2001.
Introduction to Distance Samplinn: Estimatinn abundance of biological
populations. Oxford: Oxford University Press
Burney, DA. 1980. The effects of human activity on cheetah (Acinonyx jubatus) in the
Mara region of Kenya. M. Sc. thesis. University of Nairobi, Nairobi
Carol, TM. 2000. Controversy over behavior and genetics in cheetah conservation.
Conservation Biology Series: Behavior and Conservation. 2:221-237
Carol, TM. 1994. Cheetahs of the Serenneti Plains: Group Livinn in an Asocial Species.
Chicago, London: University of Chicago Press
Carol, TM., Collins, DA. 1987. Ecological characteristics of territories of male cheetahs
(Acinonyx jubatus). Journal of Zoology, London. 21 1:89-105
Carol, TM., Laurenson, MK. 1994. Ecological and genetic factors in conservation: a
cautionary tale. Science. 263:485-486
Durant, SM. 2000. Living with the enemy: Avoidance of hyenas and lions by cheetahs in
the Serengeti. Behavioral Ecology. 11(6):624-632
Durant, SM. 1998. Competition refuges and coexistence: An example from Serengeti
carnivores. Journal of Animal Ecology. 67(3):370-3 86
Eaton, RL. 1974. The Cheetah: The Biolony, Ecolony, and Behavior of an Endannered
Species. Van Nostrand Reinhold Company, New York, NY
Emmons, LH. 1987. Comparative feeding ecology of felids in a neotropical rainforest.
Behavioral Ecology and Sociobiology. 20:271-283
Fahrig, L. 1997. Relative effects of habitat loss and fragmentation on population
extinction. Journal of Wildlife Management. 61(3):603-610
Frame, GW. 1986. Carnivore competition and resource use in the Serengeti ecosystem in
Tanzania. Ph.D. dissertation, Utah State University, Logan
Frank, LG, Woodroffe, R. 2001. Behavior of carnivores in exploited and controlled
populations. In Conservation Biology 5: Carnivore Conservation. Ed. Gittleman,
J.L., Funk, S.M., Macdonald, Wayne, R.K. University Press, Cambridge
Fritz, H, Said, S, Renaud, PC, Mutake, S, Coid, C, Monicat, F. 2003. The effects of
agricultural fields and human settlements on the use of rivers by wildlife in the
mid-Zambezi valley, Zimbabwe. Landscape Ecology. 18(3):293-302
Fuller, TK, Sievert, PR. 2001. Carnivore demography and the consequences of changes
in prey availability. In Conservation Biology 5: Carnivore Conservation. Ed.
Gittleman, JL, Funk, SM, Macdonald, Wayne, RK. University Press, Cambridge
Goss-Custard, JD, Durell, SEA. 1990. Bird behaviour and environmental planning:
approaches in the study of wader populations. Ibis. 132:272-289
Graham, A. 1966. East African Wildlife Society Cheetah Survey: Extracts from the
report by Wildlife Services. Journal of East African Wildlife. 4:50-55
Gros, PM. 1998. Status of the cheetah, Acinonyx jubatus, in Kenya: a field interview
assessment. Biological Conservation. 85(1-2):137-149
Gros, PM, Kelly, MJ, Caro, TM. 1996. Estimating carnivore densities for conservation
purposes: Indirect methods compared to baseline demographic data. Oikos.
Gros, PM, Rejmanek, M. 1999. Status and habitat preferences of Uganda cheetahs: An
attempt to predict carnivore occurrence based on vegetation structure. Biodiversity
and Conservation. 8(11): 1561-15 83
Hamilton, PH. 1986. Status of the cheetah in Kenya, with reference to sub-saharan
Africa. In Cats of the World: Biolony, Conservation, and Mananement. Ed. SD
Miller and DD Everett. National Wildlife Federation, Washington, D.C.
Hanby, JP, Bygott, JD, Packer, C. 1995. Ecology, demography, and behavior of lions in
two contrasting habitats: Ngorongoro Crater and Serengeti plains. In Serenneti II:
Research, Conservation and Mananement of an Ecosystem. Ed. ARE. Sinclair and
P. Arcese. Chicago: University of Chicago Press
Holekamp, KE, Szykman, M, Boydston, EE, Smale L. 1999. Association of seasonal
reproductive patterns with changing food availability in an equatorial carnivore, the
spotted hyaena (Crocuta crocuta). Journal of Reproduction and Fertility.
Hudak, AT. 1999. Rangeland mismanagement in South Africa: failure to apply ecological
knowledge. Human Ecology. 27(1):55-78
Hudak, AT, Wessman, CA. 1996. Textural analysis of high resolution imagery to
quantify bush encroachment in Madikwe Game Reserve, South Africa, 195 5-1996.
International Journal of Remote Sensing. 22(14):273 1-2740
IUCN. 1996. IUCN Red List of Threatened Animals. Gland, Switzerland
Janis MW, Clark, JD. 2002. Responses of Florida panthers to recreational deer and hug
hunting. Journal of Wildlife Management. 66(3):839-848
Jensen, JR. 1996. Introductory Digital Image Processing: A remote sensing perspective.
New Jersey: Prentice Hall
Karanth, KU, Nichols, JD, Kumar, NS, Link, WA, Hines, JE. 2004. Tigers and their prey:
predicting carnivore densities from prey abundance. Proceedings of the National
Academy of Sciences of the United States of America. 101(14):4854-4858
Karanth, KU, Stith, BM. 1999. Prey depletion as a critical determinant of tiger population
viability. In Ridinn the Tiner: Tiner Conservation in Human-Dominated
Landscapes. Eds. Seidensticker, J., Christie, S. and Jackson, P. Cambridge
University Press, Cambridge
Kellner, K, Bosch, OJH. 1992. Influence of patch formation in determining the stocking
rate for southern African grasslands. Journal of Arid Environments. 22(1):99-105
Kelly, MJ, Laurenson, MK, Fitzgibbon, CD, Collins, DA, Durant, SM, Frame, GW,
Bertram, BCR, Caro, TM. 1998. Demography of the Serengeti cheetah population:
the first 25 years. Journal of Zoology. 244:473-488
Kerley, LL, Goodrich, JM, Miquelle, DG, Smimov, EN, Quigley, HB, Hornocker, MG
2002. Effects of roads and human disturbance on Amur tigers. Conservation
Krauk, H. 1986. Interactions between felidae and their prey species: A review. In Cats of
the World: Biolony, Conservation, and Mananement. S. Douglas Miller and Daniel
D. Everett, editors. National Wildlife Federation, Washington D.C.
Laurenson, MK. 1995a. Implications of high offspring mortality for cheetah population
dynamics. In Serenneti II: Research, Conservation and Mananement of an
Ecosystem. Ed. ARE. Sinclair and P. Arcese. Chicago: University of Chicago
Laurenson, MK. 1995b. Cub growth and maternal care in cheetahs. Behavioral Ecology.
Laurenson, MK. 1994. The extent, timing and causes of juvenile mortality in wild
cheetahs and implications for patterns of maternal care. Journal of Zoology,
Laurenson, MK, Caro, TM, Borner, M. 1992. Female cheetah reproduction. National
Geographic Research & Exploration. 8(1):64-75
Marker, LL. Mills, MGL, MacDonald, DW. 2003. Factors influencing perceptions of
conflict and tolerance toward cheetahs on Namibian farmlands. Conservation
Marker-Kraus, L, Kraus, D. 1993. The history of cheetahs in Namibia. Swara Sep-Oct
Marker-Kraus, L, Kraus, D, Barnett, D, Hurlbut, S. 1996. Cheetah Survival on Namibian
Farmlands; the Research Findings From the CCF Farm Survey. Windhoek: Cheetah
McLaughlin, RT. 1970. Aspects of the biology of the cheetah (Acinonyx jubatus,
Schreber) in Nairobi National Park. M. Sc. thesis, University of Nairobi, Nairobi
Mech, DL. 1995. The challenge and opportunity of recovering wolf populations.
Conservation Biology. 9(2):270-278
Mills, LS, Knowlton, FF. 1991. Coyote space use in relation to prey abundance.
Canadian Journal of Zoology. 69(6):1516-1521
Milton, SJ, Dean, WRJ. 1995 South Africa's arid and semiarid rangelands: Why are they
changing and can they be restored? Environmental Monitoring and Assessment.
Moleele, NM, Perkins, JS. 1998. Encroaching woody plant species and boreholes: is
cattle density the main driving factor in the Olifants Drift communal grazing lands,
south-eastern Botswana? Journal of Arid Environments. 40:245-253
Moleele, NM, Ringrose, S, Matheson, W, Vanderpost, C. 2002. More woody plants? the
status of bush encroachment in Botswana' s grazing areas. Journal of Environmental
Morsbach, D. 1987. Cheetah in Namibia. Cat News. No. 6
Myers, N. 1975a. The cheetah's relationships to the spotted hyena: Some implications
for a threatened species. Proceedings of the 1975 Predator Symposium. Ed.
Phillilps, RL, Jonkel, C. University of Montana, Missoula
Myers, N. 1975b. The cheetah Acinonyx jubatus in Africa. IUCN monograph no. 4
O'Brien, SJ, Roelke, ME, Marker, L, Newman, A, Winkler, CA, Meltzer, D, Colly, L,
Evermann, JF, Bush, M, Wildt, DE. 1985. Genetic basis for species vulnerability in
the cheetah. Science. 227:1428-1434
O'Brien, SJ, Wildt, DE, Bush, M. 1986. The cheetah in genetic peril. Scientific
O'Brien, TG, Kinnaird, MF, Wibisono, HT. 2003. Crouching tigers, hidden prey:
Sumatran tiger and prey populations in a tropical forest landscape. Animal
Ogutu, JO, Dublin, HT. 2002. Demography of lions in relation to prey and habitat in the
Maasai Mara National Reserve, Kenya. African Journal of Ecology. 40: 120-129
Oli, MK. 1994. Snow leopards and blue sheep in Nepal: densities and predator: prey
ratio. Journal of Mammalogy. 75(4):998-1004
Poole, KG. 1994. Characteristics of an unharvested lynx population during snowshoe
hare decline. Journal of Wildlife Management. 58:608-618
Riley, SPD, Sauvajot, RM, Fuller, TK, York, EC, Kamradt, DA, Bromley, C, Wayne,
RK. 2003. Effects of urbanization and habitat fragmentation on bobcats and
coyotes in southern California. Conservation Biology. 17(2):566-576
Roelke, ME, Martenson, JS, O'Brien, SJ. 1993. The consequences of demographic
reduction and genetic depletion in the endangered Florida panther. Current
Roelke-Parker, ME, Munson, L, Packer, C, Kock, R, Cleaveland, S, Carpenter, M,
O'Brien, SJ, Pospischil, A, Hofmann-Lehmann, R, Lutz, H, Mwamengele, GLM,
Mgasa, MN, Machange, GA, Summers, BA, Appel, MJG. 1996. A canine
distemper virus epidemic in Serengeti lions (Panthera leo). Nature. 381(6578):172-
Roques, KG, O'Connor, TG, Watkinson, AR. 2001. Dynamics of shrub encroachment in
an African savanna: relative influences of fire, herbivory, rainfall and density
dependence. Journal of Applied Ecology. 38(2):268-280
Schaller, GB. 1972. The Serenneti Lion: A study of predator-prey relations. Wildlife
Behavior and Ecology Series. The University of Chicago Press, Chicago and
Sillero-Zubiri, C, Gotelli, D. 1995. Spatial organization in the Ethiopian wolf Canis
simensis-large pack and small stable home ranges. Journal of Zoology. 237:65-81
Sillero-Zubiri, C, Laurenson, MK. 2001. Interactions between carnivores and local
communities: conflict or co-existence? In Conservation Biology 5: Carnivore
Conservation. Ed. Gittleman, JL, Funk, SM, Macdonald, Wayne, RK. University
Smith, DJ. 1999. Identification and prioritization of ecological interface zones on state
highways in Florida. Proceedings of the Third International Conference on Wildlife
Ecology and Transportation. Ed. Evink, GL, Garrett, P, and Zeigler, D. Missoula,
Stander, PE, Haden, PJ, Kaqece, Ghau. 1997. The ecology of asociality in Namibian
leopards. Journal of Zoology (London). 242:343-364
Sunquist, ME, Sunquist, FC. 1989. Ecological constraints on predation by large felids.
Carnivore Behavior, Ecology, and Evolution. Ed. Gittleman, JL. Comell University
Press, Ithaca NY
Sunquist, ME, Sunquist, FC. 2001. Changing landscapes: consequences for carnivores. In
Conservation Biolony 5: Camnivore Conservation. Ed. Gittleman, JL, Funk, SM,
Macdonald, DW, Wayne, RK. University Press, Cambridge
Sutherland, WJ, Anderson, CW. 1993. Predicting the distribution of individuals and the
consequences of habitat loss: the role of prey depletion. Journal of Theoretical
Thomas, L, Laake, JL, Strindberg, D, Marques, FFC, Buckland, ST, Borchers, DL,
Anderson, DR, Burnham, KP, Hedley, SL, Pollard, JH, Bishop, JRB. 2003.
Distance 4.1. Release 2. Research Unit for Wildlife Population Assessment,
University of St. Andrews, UK.
Tobler, MW, Cochard R, Edwards, PJ. 2003. The impact of cattle ranching on large-scale
vegetation patterns in a coastal savanna in Tanzania. Journal of Applied Ecology.
Van Orsdol, KG, Hanby, JP, Bygott, JD. 1985. Ecological correlates of lion social
organization. Journal of Zoology (London). 206:97-112
White, PJ, Ralls, K. 1993. Reproduction and spacing patterns of kit foxes relative to
changing prey availability. Journal of Wildlife Management. 57(4):861-867
Woodroffe, R. 2001. Strategies for carnivore conservation: lessons from contemporary
extinctions. In Conservation Biolony 5: Camnivore Conservation. Ed. Gittleman, JL,
Funk, SM, MacDonald, DW, Wayne, RK. University Press, Cambridge
Meredith Evans received her B.A. in biology with a minor in chemistry from
California State University, Chico in December 1994. She then worked as a math and
biology teacher for 2 years in a secondary school in Malawi as a Peace Corps volunteer.
After returning home in 1997, she worked as an assistant on a variety of different
research proj ects including a study of salmon and steelhead in northern Califomnia, small
mammal diversity and abundance in Tanzania and primate demography and anti-predator
behavior in Kenya. In 1999, she j oined the Laikipia Predator Proj ect and worked as an
assistant looking at human-camnivore conflicts in the Laikipia District of Kenya. She
began her M.S. degree in the School of Natural Resources and Environment at the
University of Florida in 2001. For her research, she investigated the decline of cheetahs
in the Nakuru District of Kenya. She will begin the Ph.D. program upon completion of