ANTHRO POGENIC DISTURBANCE, ECOLOGICAL CHANGE, AND WILDLIFE CONSERVATION AT THE EDGE OF THE MARA -SERENGETI ECOSYSTEM By David S eth Green A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirement s for the degree of Zoology Ð Doctor of Philosophy Ecology, Evolutionary Biology and Behavior Ð Dual Major 2015 ABSTRACT ANTHRO POGENIC DISTURBANCE, ECOLOGICAL CHANGE, AND WILDLIFE CONSERVATION AT THE EDGE OF THE MARA -SERENGETI ECOSYSTEM By David S eth Green Biodiversity has been steadily declining in most ecosystems due to the direct and indirect effects of a growing human population. Large carnivores are particularly threatened by the negative effects of human population growth due to their slow life histories, wide -ranging behavior, and conflict with people over livestock depredation. The Maasai Mara National Reserve (henceforth, the Reserve) is located at the edge of the Mara -Serengeti ecosystem in southwestern Kenya , and is a stronghold fo r large carnivore conservation in East Africa. The Reserve has traditionally supported a great density and diversity of herbivores and large carnivore species year -round. However, current research indicates that anthropogenic activities immediately outside Reserve boundaries may be having negative effects on wildlife within the Reserve itself. My research investigates the short - and long-term effects of anthropogenic disturbance around the edges of the Reserve on wildlife populations within Reserve boundaries. First, I document longitudinal trends in the ecological and anthropogenic threats to wildlife. Next, I analyze the effects these threats have on herbivore and carnivore populations. I then test the hypothesis that spotted hyenas can serve as an indica tor species in this ecosystem. Finally, management authorities play critical roles in conserving wildlife in the Reserve, so in my last dissertation chapter, I work to clarify the effects of prescribed burn management on African mammals. My research indic ates that there have been declines in mammalian herbivore abundance and diversity from 1989 -2013 in the Talek region of the Reserve . In this same region and time period, I document significant increases in temperature, the number of pastoralist settlements , the number of livestock grazing inside the Reserve, and the number of tourist lodges. Of all these threats to wildlife, livestock inside the Reserve had the largest negative effect on native herbivores. Also, the frequency of lion sightings in the Talek region have declined by 55 % between 2004 -2008 and 2009 -2013, while the sympatric population of spotted hyenas has undergone rapid growth. These changes appear to be due to the indirect effects of a growing human population . The speed of movement exhibit ed by spotted hyenas was affected by the turning angle of their trajectory, the animalÕs proximity to anthropogenic disturbance, the time of day, the ambient temperature, the amount of rainfall, the amount of moonlight, and interaction s between anthropogen ic disturbance and social rank and anthropogenic disturbance and the time of day . The tortuosity of the paths of movement by spotted hyenas was affected by the speed of their movement, time of day, their proximity to anthropogenic disturbance , and the amou nt of rainfall . The speed at which spotted hyenas moved correlated with the diversity and abundance of resident herbivores up to 26 and 14 weeks into the future, respectively, and with the abundance of carnivores and migra tory herbivores up to 20 weeks int o the future ; these results sugges t spotted hyenas might be useful as in indicator species in this ecosystem. Prescribed burns affected the numbers of resident herbivores for 120 days, of small carnivores for 365 days, and of large carnivores for 120 days following the burning event . !iv This dissertation is dedicated to Jean and Sanford Green and Edith and David Friedman. !v ACKNOWLEDGEMENTS This research was supported by National Science Foundation ( NSF ) grants BNS8706939, BNS9021461 , IBN9309805, IBN9630667, IBN9906445, IBN0343381, IBN113170, IOB0618022, IOB0819437, IOB0920505, IOS1121474, and DEB1353110 to Kay Holekamp, and by a NSF Graduate Research Fellowship to me. I was also supported by grants from the DigitalGlobe Foundation, K enya Wildlife Trust, Sigma Alpha Mu, and the College of Natural Sciences and interdisciplinary program in Ecology, Evolutionary Biology, and Behavior at Michigan State University. I thank the Kenya National Council for Science , Technology , and Innovation and the Kenya Wildlife Service for the permissions to carry out this research. I also thank Brian Heath, the rangers and staff of the Mara Conservancy and the Narok County Governmen t, and the Senior Warden of the Maasai Mara National Reserve for their suppo rt. The staff of the Mara Serena Safari Lodge also provided me with valuable assistance in the field. The advice I received from my committee throughout my dissertation has been instrumental in my success and greatly improved this research. I would not be where I am without the significant contributions to my dissertation by Dr. Gary Roloff, Dr. Tom Getty, Dr. Catherine Lindell, Dr. Elise Zipkin, and my advisor, Dr. Kay Holekamp. I have been beyond fortunate to have an advisor that has continued to believe in me, through thick and thin. I am grateful for all that Dr. Holekamp has provided me, and for her sincere dedication to the development of her students. My research would not have been possible without her never -ending support. I am incredibly fortunate that Dr. Holekamp has allowed me to not only assist in the data collection of one of the longest !vi running carnivore monitoring projects in the world, but also to analyze a portion of her long-term data that was collected well before I joined the project. Fo r her to place this trust in me, I am extremely honored. Dr. Holekamp has also provided me with the opportunity of a lifetime to spend more than three and a half years living and working in one of the most enchanting places in the world. I want to thank Dr . Gary Roloff for his assistance in helping to design a significant portion of my field data collection. His advice during my dissertation has been extremely helpful. Living thousands of miles away from home is never easy. Fortunately, many people helped to keep my stomach full, the elephants from trampling me, and the black mambas from getting into my tent. I thank Joseph Kamaamia, Philomen Naigurian, Moses Naigurian, George Kilinyet, Jackson Kamaamia, Dickson Pion, Stephen, and Lesingo for taking extreme ly good care of me in both fisi camps. Their familial treatment made me feel like I was not just comfortable in a place far from home, but rather, living in a second home that I had not previously realized existed. Many people in Kenya helped to make the u nexpected events of living in the field not only bearable, but also fun. I thank Amanda Subalusky, Chris Dutton, Stephanie Dloniak, Howard Saunders, J.P. and Terilyn Lemaire, Milton and Elly Kirkman, William Deed, Larijin Lakakwar, and Wa and Daniel Beckwi th, for saving me from spending countless nights in the bush, for forcing me to take breaks, and for being there when I needed them the most. The countless members of the Holekamp lab and the Mara Hyena Project have provided me with camaraderie unlike any other in Kenya and in East Lansing. Far too !vii numerous to name everyone, I thank Wes Binder, Andy Booms, Katy Califf, Hadley Couraud, Leslie Curren, Andy Flies, Andy Gersick, Julia Greenberg, Lily Johnson -Ulrich, Sarah Jones, Wilson Kilong, No ”mie Lamon, Zac h Laubach, Kenna Lehmann, Nora Lewin, Kevin McCormick, Tracy Montgomery, Jenna Parker, Benson Pion, Phoebe Parker -Shames, Kate Shaw, Eli Strauss, Eli Swanson, Emily Thomas, Emily Thorne, Julie Turner, and Meg Wallen for providing excellent assistance in th e field and/or Michigan. While at Michigan State, I also had the privilege of working with many stellar undergraduate researchers, including Emily Bernhardt, Shannon Carvey, Mary Kearney, John Kochiss, Morgan Lucot, Ellen McGowan, and Taylor Pederson. I tr uly appreciate all of their dedication to my dissertation. Having a good support network of friends was imperative to my success in Michigan. I want to thank Trevor and Megan Meckley, Austin Dreyer, Kari Dammerman, Rachael Eaton, Michelle Lute, Sonja Chris tensen, Jeff Smith, Rohan Maddamsetti, Scott and Sarah Resnick, Shira Wolf, Marc Wiseman, and Siri Okamoto for all of their support in Michigan and elsewhere. Finally, and most importantly, I want to thank my family for always believing in me and encouragi ng me to follow my dreams. I know my parents were not always the happiest when this encouragement had led me, their youngest son, to be living halfway around the world. However, they have always been there for me and had confidence in the path that I chose for my career. My brothers, grandparents, and extended family have also been a constant source of unwavering support and love. For all of this, thank you. !viii TABLE OF CONTENTS LIST OF TABLES !!!!!!!!!!!!!!!!!!!!!!!!!!!.!... x LIST OF FIGURES !!!!!!!!!!!!!!!!!!!!!!!!!!!!.. xii CHAPTER 1: General introduction !!!!!.. !!!!!!!!!!!!!!!!. 1 Introduction !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!...2 Summary of this dissertation !!... !!!!!!!!!!!!!!!!!!!!..5 Writing style of this dissertation !!!!!!!!!!!!!!!!!!!!!!.6 CHAPTER 2: Long-term ecological change in the eastern Maasai Mara National Reserve !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!..!!... 7 Introduction !!!!!!!!!!!!!!!!!!!!!!!!...!!!.!!.. 8 Methods !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! 11 Study area !!!!!!!!!!!!!!!!!!!!!!!!!!!!!! 11 Data collection !!!!!!!!!!!!!!!!!!!!!!!!!!!! 11 Herbivore data !!!!!!!!!!!!!!!!!!!!!!!!!!! 11 Meteorological data !!!!!!!!!!!!!!!!!!!!!!!!! 14 Number of pastoralist settlements !!!!!!!!!!!!!!!!!!... 14 Livestock !!!!!!!!!!!!!!!!!!!!!!!!!!!!!.. 15 Tourism !!!!!!!!!!!!!!!!!!!!!!!!!!!!!... 15 Statistical analyses !!!!!!!!!!!!!!!!!!!!!!!!!!. 15 Results !!!!!!!!!!!!!!!!!!!!!!!!!!!!.!!!!. 19 Herbivore abundance and diversity !!!!!!!!!!!!!!!!!!!. 19 Meteorological data !!!!!!!!!!!!!!!!!!!!!!!!!! 23 Number of pastoralist settlements !!!!!!!!!!!!!!!!!!!... 26 Livestock !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!.. 26 Tourism !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! 28 Hypothesis testing !!!!!!!!!!!!!!!!!!!!!!!!!!.. 30 Discussion !!!!!!!!!!!!!!!!!!!!!!!!!.!!!!!... 32 CHAP TER 3: Effects of anthropogenic disturbance on long -term trends in the demography and behavior of lions and spotted hyenas in the Maasai Mara National Reserve, Kenya !!!!... !!!!!!!!!!!!!!!!!!...!!!!!!. 40 Introduction !!!!!!!!!!!!!!!!!!!!!!!!!!.!!!!. 41 Methods!!!!!!!!!!!!!!!!!!!!!!!!!!!!!.!!... 46 Study site !!!!!!!!!!!!!!!!!!!!!!!!!!!!!! 46 Data collection !!!!!!!!!!!!!!!!!!!!!!!!!!!!. 49 Demography and behavior of spotted hyenas !!!!!!!!!!!!!.. 49 Estimating lion abundance and documenting their us e of space !!!!!... 51 Monitoring of ecological and anthropogenic variables !!!!!!!!!! 51 Statistical analyses !!!!!!!!!!!!!!!!!!!!!!!!!! 52 Longitudinal trends in size of the Talek West clan and comparisons with other clans in the Reserve !!!!!!!!!!!!!!!!!!!!!!!!.. 53 !ix Trends in lion interactions with hyenas and lion sightings !!!!!!!!! 53 Spatial analysis of hyena and lion locations !!!!!!!!!!!!!!! 54 Results !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!.. 55 Temporal trends in the clan sizes of spotted hyenas !!!!!!!!!!!! 55 Trends in lion interactions with hyenas and lion sightings !!!!!!!!!... 57 Spatial analysis of hyena and lion locations !!!!!!!!!!!!!!!.. 60 Discussion !!!!!!!!!!!!!!!!!!!!!!!!!!!!..!!.. 62 CHAPTER 4: Early warning signals of environmental degradation : Do the movements of spotted hyenas indicate the diversity and abundance of sympatric wildlife? ............... 67 Introduction !!!!!!!!!!!!!!!!!!!!!!!!!!.!!!!. 68 Methods !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! 72 Study site !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!. 72 Monitoring the movements by spotted hyenas !!!!!!!!!!!!!!.. 75 Sampling the abundance of sympatric wildlife !!!!!!!!!!!!!!... 78 Statistical analyses !!!!!!!!!!!!!!!!!!!!!!!!!!. 80 Predictors of movement patterns by spotted hyenas !!!!!!!!!!... 80 Spotted hyenas as an indicator species !!!!!!!!!!!!!!!!. 82 Results !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!.. 84 Predictors of movement patterns by spotted hyenas !!!!!!!!!!!! 84 Spotted hyenas as an indicator species !!!!!!!!!!!!!!!!!. 88 Discussion !!!!!!!!!!!!!!!!!!!!!!!!!!!!.!!... 94 CHAPTER 5: Temporal dynamics of the responses by African mammals to prescribed fire .!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!... !!!.. 99 Introduction !!!!!!!!!!!!!!!!!!!!!!!!!!!!.!!. 100 Study area !!!!!!!!!!!!!!!!!!!!!!!!!!!!!....... 103 Methods !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! 106 Line-transect sampling !!!!!!!!!!!!!!!!!!!!!!!!.. 106 Statistical analyses of herbivore data !!!!!!!!!!!!!!!!!!. 108 Impact of prescribed burning on carnivore visibility !!!!!!!!!!!!.. 111 Estim ating vegetation recovery with the Normalized Difference Vegetation Index (NDVI) !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!.. 111 Results !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!.. 112 Effects of season, transect location, and precipitation on herbivore densities !. 112 Effects of burning on herbivores !!!!!!!!!!!!!!!!!!!!.. 115 Temporal dynamics of burn effects on herbivores !!!!!!!!!!!!! 116 Impact of burning on carnivore visibility !!!!!!!!!!!!!!!!!.. 117 Vegetation recovery with NDVI !!!!!!!!!!!!!!!!!!!!! 120 Discu ssion !!!!!!!!!!!!!!!!!!!!!!!!!!!!..!!.. 121 Management implications !!!!!!!!!!!!!!!!!!!!!!!!.. 126 LITERATURE CITED !!!!!!!!!!!!!!!!!!!!!!.!!!!!. 128 !x LIST OF TABLES Table 2.1. Mean posterior parameter estimates (and 95 % credible intervals) of random and fixed effects from the hierarchical models that investigated trends in the numbers of resident and migrant herbivores, richness, and evenness from 1989 -2013 in the Talek region of the Maasai Mara National Reserve. Significant posterior estimates, not including the intercept, are indicated in bold font (95 % credible intervals not overlapping zero). Month of the year was modeled as a random effect on the intercept; here we present the average for all 12 months. Parameter estimates are for the transect loc ated in long-grass habitat !!!!!!!!!!!!!!!!!!!!!!!!!!!!.21 Table 2.2. Mean posterior parameter estimates and 95 % Credible Intervals (95 % CI) of the models investigating the temporal trends in ecological and anthropogenic factors in the Talek region. A ne gative binomial distribution with a logit link was used for livestock, and a Poisson distribution with a log link was used for the number of bomas, tourist lodges, and beds for tourists; all other analyses were performed using a normal distribution and an identity link. All posterior parameter estimates are presented on the normal scale. The parameter Year indicates the amount of change in each subsequent year. Significant posterior parameters are indicated in bold font (95 % CI not overlapping 0) !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!.. 24 Table 2.3. Mean posterior parameter estimates (and 95 % credible intervals ) from the hierarchical models that tested hypotheses to explain the variation in herbivore diversity and abundance in the Talek region. Statistically sign ificant parameter estimates are highlighted in bold font (95 % credible intervals that did not overlap zero). Month of the year was modeled as a random effect on the intercept ; here we present the average for all 12 months. Parameter estimates are for the transect located in long -grass habitat!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! 31 Table 3.1. Summary of the mean posterior parameter estimates and 95 % Credible Intervals (CIs) for longitudinal trends in size of the Talek West clan of spotted hyenas (1988 -2013), and cross comparisons between the Talek West clan and other clans in the Reserve between 2002 and 2013. The Talek West clan has increased significantly over time, and was significantly larger than all other study clans. Significant posterior parameter es timates are indicated in bold font (95% CIs that did not overlap with zero). The parameter Year indicates the amount of change in each subsequent year !!.. 57 Table 3.2. Mean posterior summaries and 95 % Credible Intervals (CIs) for variables explaining the distances in meters at which spotted hyenas and lions were found from the Reserve boundary over time in the Talek region of the Maasai Mara National Reserve, Kenya. Locations of hyenas and lions considered here were recorded from 2004-2013. Variables are listed in descending order of importance following year and hour based on the weights of their untransformed posterior parameter estimates. Statistically significant posterior parameter estimates are indicated in bold font (95% CIs that did not overlap wi th zero) !!!!!!!!!!!!!!!!!!!!!!!! 61 !xi Table 4.1. Salient characteristics of the Talek West, Serena North, and Serena South clans of spotted hyenas monitored in the current study in 2013. Standard errors are presented in parentheses. The number of tourists p er territory was estimated as the number of vehicles counted biweekly during counts by Mara Hyena Project personnel !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!.. 77 Table 4.2. Results from the generalized linear mixed -effects models that predicted the speed and turni ng angles of spotted hyenas between successive locations separated by one hour inside and around the Maasai Mara National Reserve, Kenya. Parameter estimates, Standard Errors (SEs), Degrees of Freedom (DF), T-, and P-values are displayed. Significant param eter estimates are indicated in bold font (P < 0.05) !!. 87 Table 4.3. Results from the generalized linear mixed -effects models that predicted herbivore and carnivore abundance, and species richness, as a function of the speed, turning angles, and rank of spotted hyenas in the same area previous to the count. The behaviors of spotted hyenas significantly predicted the numbers of resident herbivores 2, 8, and 14 weeks before the sample, species richness 2, 8, 14, 20, and 26 weeks before the sample, and numbe rs of migrant herbivores and carnivores 2, 8, 14, and 20 weeks before the sample. Significant parameter estimates are indicated in bold font (P < 0.05)!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!.. . 92 Table 5.1. The results from the ANOVA F-test and probabilities derived from generalized linear models that predicted herbivore densities based on ecological factors for 8 herbivore species in the Mara -Serengeti ecosystem, southwest Kenya, 2008 -2011. Bolded P values indicate statistically significant effects. Denominat or df = 168 in all global models!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!... 113 Table 5.2. Results from the ANOVA F-test and probabilities derived from generalized linear models that predicted herbivore densities based on burn status in the Mara -Serengeti ecosystem , southwest Kenya, 2008 -2011. df = 1,168 in all comparisons. .. 116 !xii LIST OF FIGURES Figure 2.1. The Talek region of the Maasai Mara National Reserve in southwestern Kenya. The locations of pastoralist settlements ( o) and permanent tourist lodges and cam ps ( !) in (a) 1991 and (b) 2012 within 2 km of the Reserve border in the Talek region. The solid line is the boundary of the Reserve, and the shaded area lies within Reserve boundaries. All pastoralist settlements within 2 km of the Reserve boundary (dotted li ne) are noted. (c) The location of the Talek region in the Maasai Mara National Reserve (dashed lines) and the community conservancies north of the Reserve (dark shaded regions). The two 4 -km transects along which we monitored herbivores at biweekly interv als in the current study are noted with solid lines !!!!!!!!! 13 Figure 2.2. Temporal trends in the modeled count s of (a) resident and (b) migrant herbivores, (c) species richness, and (d) the Shannon index in the Talek region by month between 1989 -2013. Mean ± 95 % credible intervals are displayed !!!!.. 20 Figure 2.3. Temporal trends in the mean expected yearly count of (a) resident and (b) migrant herbivores, (c) species richness, and (d) the Shannon index in the Talek region between 1989-2013. The trans ect in short grass habitat is indicat ed with solid lines and the transect in long grass habitat is indicat ed with dashed lines (a-c). We present the average for both transects because there was no statistical difference in the Shannon index between transec ts (d). All declining trends were significant (95 % credible intervals did not overlap zero for the parameter of trend over time )!!!!!!!... 22 Figure 2.4. Yearly temporal trends in (a) the drought index in the Talek region of the Maasai Mara National Rese rve, and (b) mean ± standard error ( SE) maximum and (c) minimum temperatures in Narok, Kenya. Solid lines are the polynomial equations fit to the data using general linear and generalized linear mixed -effects models in R , and dashed lines indicate the 95 % credible interval around this equation. The drought index is the number of standard deviations from the mean precipitation for the entire study where a given yearÕs precipitation was observed to fall !!!!!!!!!!!!! 25 Figure 2.5. Temporal trends in the (a) total number of pastoralist settlements along the edge of, and (b) mean numbers (± standard error) of livestock grazing daily inside, Reserve boundaries in the Talek region between 1988 -2013. Solid lines represent the polynomial equations fit to the data using general linear and generalized linear mixed -effects models in R , and dashed lines indicate the 95 % credible interval around this equation!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! 27 Figure 2.6. Temporal trends in the number of tourist lodges ( ") and beds a vailable to tourists ( o) in the Talek region of the Maasai Mara National Reserve, Kenya, between 1988-2013. Solid lines represent the polynomial equations fit to the data using general linear models in R , and dashed lines indicate the 95 % credible interva l around this equation!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! 29 !xiii Figure 2.7. The relative importance of the ecological and anthropogenic variables affecting the abundance of resident and migrant herbivores, species richness, and Shannon index. Mean posterior es timates and 95 % Credible Intervals (CIs) are shown !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! 32 Figure 2.8. The intensive grazing taking place in the Talek region. (a) Multiple herds like this graze inside the Reserve daily. (b) The effects of intensive, sustain ed grazing in the Talek region on vegetation and soils, as seen from a satellite orbiting the earth in space (Photo credit: Google Earth). The near -vertical swaths across this landscape are the tracks left by large cattle herds moving back and forth into t he Reserve from bomas situated outside the northern Reserve boundary, indicated here by the jagged line of riverine vegetation !!!!!!!!!!!!!!!!!!!!!!!!!!!!... 35 Figure 3.1. Sites of carnivore monitoring in the Maasai Mara National Reserve, Kenya. The approximate boundaries of the six territories defended by members of our study hyena clans are indicated with dashed lines. Study clans in the Mara Triangle (shaded region), which is managed by the Mara Conservancy, are Serena North (SN), Serena South (SS), an d Happy Zebra (HZ). Clans in regions managed by the Narok County Government (no shading) are Talek West (TW), Fig Tree (FT), and Mara River (MR). The TW clan is located in the Talek region of the Reserve. The star (!) indicates the closest town bordering t he Talek region (the town of Talek) !!!!!!!!!!! 48 Figure 3.2. Trends in the sizes of spotted hyena clans in the Maasai Mara National Reserve, Kenya. The yearly mean ± standard error (SE) clan sizes are presented for the Talek West (1988 -2013), Fig Tree (2 007-2013), Mara River (2002 -2013), Serena North (2008 -2013), Serena South (2008 -2013), and Happy Zebra clans (2008 -2013). Years without SE bars did not have multiple censuses. The two documented clan fission events in Talek West that occurred in 1989 and 2 000 are indicated with horizontal lines!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! 56 Figure 3.3. The expected mean number of lions sighted per month per clan of hyenas ± 95 % credible interval ( CI) between 2004-2008 and 2009-2013. The territory of the Talek West clan (TW) is located in the Talek region, and the territories of the Serena North (SN), Serena South (SS), and Happy Zebra (HZ) clans are located in the Mara Triangle. The expected mean count in TW significantly declined between 2004 -2008 and 2009-2013 (p osterior parameter estimates 95 % CIs did not overlap with one another). The expected mean monthly count of lions in TW in 2009 -2013 was significantly lower than SN, but no different than SS or HZ, during the same time period !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!. 59 Figure 3.4. Mean ± standard error (SE) distances at which spotted hyenas and lions were found from the Reserve boundary (a) over years of the study, and (b) over hours of the day during which we monitored lions and spotted hyenas !!!!!!!!! 62 Figure 4.1. Locations of the study clans and loops (dotted lines) run to sample sympatric wildlife in the Maasai Mara National Reserve, Kenya. The clans in areas managed by !xiv the Mara Conservancy are Serena North (SN) and Serena South (SS). The clan of hyenas in the region exposed to severe anthropogenic disturbance, and managed by the Narok County Government, is the Talek West clan (TW). Talek West is located in the Talek region of the Reserve. Territory boundaries are indicated with thick dashed lines. The loca tions of the weather stations ( ) and the town (!) closest to the TW territory (Talek) are also noted !!!!!!!!!!!!!. !!!!!!!!!!. 74 Figure 4.2. Speeds of movement by spotted hyenas inside Reserve boundaries of the Maasai Mara National Reserve, Kenya, in are as managed by the Narok County Government ( ) and the Mara Conservancy ( !). (a) Mean ± standard error ( SE) speeds by individuals throughout the night. Hours of darkness are shaded in grey. (b) Mean ± SE speed of movement by spotted hyenas by rank and manag ement agency !!! 86 Figure 4.3. Variation in the abundance and distribution of (a) resident (b) and migrant herbivores, (c) carnivores, and (d) species richness within the Serena North (SN), Serena South (SS), and Talek West (TW) territories in the Maasai Mara National Reserve, Kenya. Darker shades indicate a comparatively greater abundance (a -c), or diversity (d) in that 200m x 200m grid cell than in lighter cells during this study, accounting for effort. Open areas within territory boundaries indicate reg ions with no sampling effort !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!.. 89 Figure 5.1. Herbivore and carnivore survey transect locations in the M aasai Mara National Reserve, southwestern Kenya, 2008 -2011. The four burn transects monitored in the Reserve are shown wi th the dates on which burns were initiated and the sample size representing the total number of times each transect was run . The dark triangle indicates the location of our weather station !!!!!!!!!!!!!!!!!... 105 Figure 5.2. Detection probabilities for th e 8 herbivore species monitored in this study based on unpu blished data from D. S. Green. Detection distances were calculated using a rangefin der during June -August of 2011. Solid lines indicate the detection function for animals in burned areas, whereas d ashed lines indicate the detection function for ani mals seen in non -burned areas. Figures with only one line indicate no differences between detection distances as a function of burned status. The vertical line indicates the maximum distance for which anim als were recorded in the study, and the horizontal line indicates the 90 % detection probabilities in burned and unburned areas !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!.. 107 Figure 5.3 . Density of ThomsonÕs gazelle and topi on burned and unburned areas, relativ e to the time of the fire, in the Ma asai Mara National Reserve . ÒPreburnÓ includes averaged data collected prior to the burn . Each subsequent interval on the x -axis represents a 30 -day block of time after the burn . Error bars represent 95% confidence intervals around the means . Two asterisks indicate significant differences between densities on burned vs. unburned areas at P # 0.001; one asterisk indicates significant differences at P < 0.05 !!!!!!!!!!!!!!!!!!!!!!!!!!! 117 !xv Figure 5.4 . Carnivore abundance in burned areas, relative to the time of the fire, in the Maasai Mara National Reserve . Two asterisks indicate significant differences from preburn abundances at P # 0.001; a single asterisk indicates significant differences from preburn abundances at P < 0. 05 (!2 test of independence with YatesÕ correction) . Error bars represent 95% confidence intervals around the means . Sample sizes for each grouping, representing the number of transects run, appear below the bars !!!! 119 Figure 5.5 . The response of local vegetation to prescribed burning assessed using the Normalized Difference Vegetation Index (NDVI) . The values represented here are the averaged monthly values for the 2 ~ 5 x 5 km grid cells in which > 60% of the burn transects fell for the prescribed burn s taking place on 11 Jul 2009 and 27 Jun 2010 . Values were grouped by their Days Since Burn (DSB), and ÒPreburnÓ indicates the 30 -day interval before burning was initiated . Error bars indicate 95% confidence intervals around the means. ....................................................................................................... 120 1 CHAPTER 1 : General introduction 2 Introduction The conservation outlook for much of the worldÕs wildlife is currently very grim. Of the more than 7.5 million esti mated species of animals on the planet (Mora et al. 2011) , the International Union for the Conservation of Nature (IUCN) currently lists over 20% of those that have been described as threatened with extinction (IUCN 2015) . Alt hough the background rate of extinction is estimated to be two species per 10,000 species per 100 years, some researchers estimate that we may currently be experiencing rates of extinctions up to 100 times greater than this (Dirzo et al. 2014; Ceballos et al. 2015) . The world is in the midst of a mass extinction event, the sixth in the history of the planet, and one that is being caused primarily by effects associated with human population growth (McKee et al. 2003; Butchart et al. 2010; IUCN 2015) . Of the wildlife species that are most threatened with e xtinction in our lifetimes, large mammalian carnivores are a particularly sensitive group given their slow life histories , wide ranging behavior, and conflict with people over personal safety and livestock depredation . The reduction or complete extirpation of large mammalian carnivores can have ecosystem -wide consequences due to the reduction or remova l of the top-down control many large carnivores exert over species at lower trophic levels in natural ecosystems (Estes et al. 2011; Ripple et al. 2014) . The direct threats of human population growth on large mammalian carnivores often stem from conflict over livestock depredation, and can result in the mortality of large carnivores (Woodroffe 2001; Woodroffe & Frank 2005; Ripple et al. 2014; Ogada 2014). The indirect threats of a growing human population to mammalian carnivore populations have been harder to elucidate because they may take years to manifest. However, they can include decreased numbers of prey 3 from overharves ting, increased levels of vigilance behavior (e.g., Pangle & Holekamp 2010), shift s in activity patterns (e.g., Van Dyke et al. 1986; Frank & Woodroffe 2001; Carter et al. 2012; Rasmussen & Macdonald 2012) , and altered stress physiology (e.g., Creel et al. 2002; Van Meter et al. 2009; Creel et al. 2013b; Bhattacharjee et al. 2015) . Whereas management and conservation efforts have been implemented to curb and halt the direct effects of a grow ing human population on carnivores, both its indirect effects and techniques to manage the se effects remain poorly understood . The Mara -Serengeti ecosystem in East Africa is a stronghold for the conservation of large carnivores (Riggio et al. 2013) , it is a n important source of foreign exchange for local and national economies (Karanja 2003; Norton -Griffiths et al. 2008; Polasky et al. 2008), and it is a challenging area to manage because of its myriad stakeholders. The northernmost portion of the Mara -Serengeti ecosystem is the Maasai Mara National Reserve (henceforth, the Reserve) . The Reserve is r eferred to as the ÒJewel in KenyaÕs crownÓ because of its spectacular wildlife viewing (Walpole et al. 2003) ; the Reserve is host to a remarkable density and diversity of herbivore and carnivore species throughout the year . These resident species are joined seasonally b y an estimated 1.2 million wildebeest (Connochaetes taurinus ) and hundreds of thousands of zebra (Equus quagga; Bell 1971; Maddock 1979; Sinclair & Arcese 1995) . Unfortunately , there are various threats to wildlife in the Reserve. Habitat change and livestock grazing inside and outside of the Reserve may be responsible for the widespread declines in herbivore populations within the Reserve (Ottichilo et al. 2000; Ogutu & Owen -Smith 2005; Ogutu et al. 2005; 2009; 2011) . Anthropogenic disturbance inside the Reserve has a lso been closely associated with behavioral (Boydston et al. 4 2003a; Kolowski et al. 2007; Kolowski & Hole kamp 2009; Pangle & Holekamp 2010) , physiological (Van Meter et al. 2009) , and demographic changes in spotted hyenas (Crocuta crocuta ; Watts & Holekamp 200 9; Pangle & Holekamp 2010) , the most abundant large mammalian carnivore in this ecosystem. Limited information is available on how anthropogenic disturbance may be affecting other large carnivores in this ecosystem, such as lions (Panthera leo ; but see Ogutu et al. 2005) . In the Reserve , we are presented with a unique situation that allows us to further understand how human activities are affecting wildlife populati ons. The eastern portion of the Reserve is managed by the Narok County Government. Wildlife species in this region are exposed to anthropogenic disturbance by livestock grazing inside the Reserve, direct conflicts with people over livestock losses, and an unregulated tourism industry . In contrast to the east, the western portion of the Reserve has been under strict management by the Mara Conservancy since the early 2000Õs (Walpole & Leader -Williams 2001) . The Mara Conservancy restricts livestock grazing, limits tourism development, and maintains a relatively pristine environment for wildlife . This range of human activity within the same ecosystem creates the rare opportunity for a natural experiment to investigate the effects of anthropogenic disturbance on wild life. In my dissertation, I use this naturally -occurring variation in anthropogenic disturbance to inquire how herbivore and carnivore populations within the Reserve are faring across time and space. I also investigate ways in which conservation and management organizations can best conserve the populations of wildlife that are declining. This research represents a significant contribution to conservation and management efforts in the region. 5 Summary of this d issertation In 1988, Drs . Kay Holekamp and Laur a Smale began an intensive study of the behavioral ecology of spotted hyenas in the Reserve (the Mara Hyena Project) . From 1988 to the present (27 years), The Mara Hyena Project has monitored the behavior, demography, and physiology of up to 6 clans of spo tted hyenas, the abundance and diversity of mammalian herbivores, the numbers and spatial distributions of sympatric carnivore species in the Reserve, and the ecological and anthropogenic factors that may be affecting local wildlife populations (e.g., dail y temperature and rainfall, numbers of livestock grazing within Reserve boundaries). Long -term datasets like these are imperative for asking questions related to population trends over time. I utilize this long -term dataset, as well as new data I collected , to investigate how anthropogenic disturbance and ecological change are affecting wildlife populations in both the short - and long-term within the Reserve. I begin my dissertation, in Chapter 2, by characterizing the long -term effects of human population growth and anthropogenic disturbance on herbivore populations in the Reserve . In Chapter 3, I investigate the long-term effects of anthropogenic disturbance on the populations and behavior of large carnivore s inside the Reserve. In Chapter 4, I inquire whe ther the behavior of spotted hyenas can predict declines in populations of sympatric wildlife. In my final chapter, Chapter 5, I attempt to clarify the temporal responses of African mammals to grassland management by prescribed burn ing. The overarching goa l of my dissertation is to clarify the effects of anthropogenic disturbance around the Reserve on wildlife communities living within Reserve 6 boundaries. My dissertation presents an integrative approach that incorporates behavioral and demographic data to understand the effects of human population growth on wildlife in an important protected area in Kenya . The results from my dissertation provide information critical for understand ing the long-term effects of anthropogenic disturbance on wildlife populations . Writing style of this dissertation My dissertation is the culmination of work done for many years before I started my own research, and collaborations with other scientists . I am extremely grateful for all of their contributions to this dissertation , and they will co -author all publications with me presenting the work described here ; I will therefore use the first person plural throughout the remainder of this dissertation to recognize their contributions. 7 CHAPTER 2: Long-term ecological change in the eastern Maasai Mara National Reserve 8 Introduction Worldwide human population growth is a major threat to biological diversity (McKee et al. 2003; Butchart et al. 2010) . Protected Areas (PAs) represent an important mechanis m for wildlife conservation, particularly in developing nations, and numbers of PAs have been increasing over time (Butchart et al. 2010) . In sub -Saharan Africa, the fragmentation of wildlife habitat in an expanding matrix of agriculture and urbanization has only amplified the importance of PAs for conserving critical habitat (Newmark 2008) . However, i ntensive anthrop ogenic activity near PAs may exacerbate threats to biodiversity within PA boundaries (Woodroffe & Ginsberg 1998; Wittemyer et al. 2008; Newmark 2008; Craigie et al. 2010) . There is now a critical need to under stand exactly how anthropogenic activity around the edges of PAs affects wildlife populations within park boundaries. Unfortunately, these effects are difficult to document because they may take many years to emerge, and long -term ecological datasets are c hallenging and expensive to generate and maintain. One of the most important PAs in East Africa in terms of its extraordinary biodiversity, the observability of its wildlife, and its ability to attract both tourists and foreign exchange, is the Maasai Mara National Reserve (henceforth, the Reserve) in southwestern Kenya. The Reserve, situated in the northernmost portion of the Mara -Serengeti ecosystem, is host to a remarkable diversity of wildlife. In addition to supporting high densities of resident herbiv ores and carnivores, it also serves as the northernmost destination for the annual migrations of wildebeest ( Connochaetes taurinus ) and zebra ( Equus quagga ) from Serengeti National Park (Bell 1971; Maddock 1979; Sinclair & Arcese 1995) . The Reserve is contiguous with the Serengeti National 9 Park to the south, but flanked by expand ing populations of Maasai pastoralists along all of its other borders. Since the 1950s, rapid human population growth along the edges of the Reserve has led to loss of habitat suitable for wildlife, increased levels of mechanized agriculture, and heightene d grazing pressure from livestock (Serneels et al. 2001; Thompson & Homewood 20 02; Ogutu et al. 2005; 2009). It has been hypothesized that these changes in habitat might be contributing to declines in numbers of many resident herbivores within and around the Reserve (Ottichilo et al. 2000; Serneels & Lambin 2001; Lamprey & Reid 2004; Ogutu et al. 2005; 2009; 2011) . Monitoring efforts started in 1960 have documented declines in resident wild herbivores at levels exceeding 50 % for some species, both inside and outside Reserve boundaries (Ottichilo et al. 2000; Homewood et al. 2001; Serneels & Lambin 2001; Ottichilo et al. 2001; Ogutu et al. 2011). Numbers of migra nt herbivores visiting the Reserve have also been declining; since 1984, there has been a 35 % decrease in the numbers of migratory wildebeest that visit the Reserve each year (Norton -Griffiths et al. 2008) . Conflict between people and wildlife in and around the Reserve is a complex issue compounded by the changing lifestyles of the loc al Maasai pastoralists. Once nomadic, Maasai pastoralists around the Reserve have become sedentary in recent decades (Ogutu et al. 2009) , and are investing in agriculture and livestock production around Reserve boundaries (Homewood et al. 2001; Thompson & Homewood 2002; Thompson et al. 2002; Norton -Griffiths et al. 2008) . The presence of increasing numbers of pastoralists around Reserve boundaries may al so be influencing wildlife populations. Overall numbers of livestock in the region are poorly documented, but are believed to be stable or to have increased since the 1970s (Norton -Griffiths et al. 2008; 10 Ogutu et al. 2011), with their numbers only declining in years of drought (Ottichilo et al. 2000). Furthermore, the subdivision of communally -owned lands around the Reserve and lax regulation of Rese rve rules have encouraged pastoralists to rely on illegally grazing livestock within park boundaries year -round (Boydston et al. 2003a; Kolowski & Holekamp 2009; Ogutu et al. 2009; Butt 2014) , especially during drought years. In addition to livestock grazing, a poorly regulated tourism industry may also be having adverse effects on resident wildlife near Reserve boundaries. Hundreds of thousands of tourists visit the Reserve each year . General disregard for a formal moratorium on the cons truction of new tourist facilities has resulted in many new lodges being built along the edges of the Reserve in recent years (Karanja 2003; Reid et al. 2003). Whether the presence of these tourist lodges affects nearby wildlife is unknown. Using data collected from 1988 -2013, our goal was to investigate the effects of anthropogeni c disturbance and ecological change on wild mammalian herbivores in the Talek region of the Reserve. To do this, we documented long -term trends in herbivore diversity and abundance, and simultaneously monitored natural ecological and anthropogenic variable s that might be affecting these measures. We then used hierarchical models to test hypotheses suggesting several possible explanations for observed temporal patterns in herbivore abundance and diversity . These hypotheses suggest that herbivore abundance an d diversity might be affected by rainfall patterns, increasing temperature, increasing numbers of pastoralists living near Reserve borders, the expansion of tourist facilities along Reserve boundaries, numbers of livestock grazing within Reserve boundaries , and the interaction between livestock grazing and rainfall. Although previous researchers have documented declines in wild herbivore 11 populations in the Reserve (Ottichilo et al. 2000; Ogutu et al. 2009; 2011) , none have yet to investigate trends in overall biodiversity, or explicitly test hypotheses invoking relationships between ecological and anthropogenic effects on herbivores simultaneously (but see Ogutu et al. 2009) . Methods Study area The Reserve is primarily comprised of open grassland interspersed with riparian areas. It has traditionally supported large herds of resident and seasonally migrant herbivores alongside populations of small and large carnivores throughout the year (Bell 1971; Sinclair & Norton -Griffiths 1979; Stelfox et al. 1986; Craft e t al. 2015). Rainfall patterns are bimodal, with most rain falling in November -December and March -May. Because the study site is less than 2 degrees south of the equator, there is limited seasonal variation in temperature. Our study took place in the Tale k region (henceforth, Talek), located in the northeastern portion of the Reserve (Figure 2.1). The Talek region is approximately 143 km 2 in size, and is characterized by a flourishing tourism industry and exponential human population growth along the edges of the Reserve (Figure 2.1a,b; Reid et al. 2003; Lampre y & Reid 2004) . Data collection Herbivore data Two 4-km line -transects were surveyed biweekly in Talek starting in October 1988; further detail on methods used for counting herbivores can be found in previous 12 publications (Ho lekamp et al. 1999; Boydston et al. 2003a; Van Meter et al. 2009) . The northernmost transect was located in short -grass habitat, while the southernmost transect was located in long -grass habitat (Figure 2.1c). Resident herbivores monitored in the current study were impala (Aepyceros melampus) , topi (Damaliscus korrigum jimela ), CokeÕs hartebeest ( Alcelaphus buselaphus cokii), warthog ( Phacochoerus aethiopicus ), GrantÕs gazelle ( Gazella granti ), cape buffalo ( Syncerus caffer ), hippopotamus ( Hippopotamus amp hibius), Maasai giraffe ( Giraffa camelopardalis tippelskirchi ), eland ( Taurotragus oryx ), duiker ( Sylvicapra grimmia ), African elephant (Loxodonta africana ), oribi ( Raphicerus campestris ), reedbuck ( Redunca redunca ), waterbuck ( Kobus ellipsiprymnus ), bushb uck ( Tragelaphus scriptus ), and ThomsonÕs gazelle (Eudorcas thomsoni ). ThomsonÕs gazelle in Talek are non -migratory and present year -round; here they were considered to be residents in all analyses. The migra nt herbivores we monitored were wildebeest and z ebra. We also analyzed two metrics to investigate trends in mammalian herbivore biodiversity: species richness and ShannonÕs diversity index of evenness (henceforth, evenness). We did this to look at both the total biodiversity in Talek (richness), as well as the weighted abundance of each species in the herbivore community (evenness). We calculated species richness as the total number of different herbivore species sighted on each transect, and evenness was calculated for all transects on which more than o ne species were seen. 13 Figure 2.1. The Talek region of the Maasai Mara National Reserve in southwestern Kenya. The locations of pastoralist settlements ( o) and permanent tourist lodges and camps ( !) in (a) 1991 and (b) 2012 within 2 km of the Reserve border in the Talek region. The solid line is the boundary of the Reserve, and the shaded area lies within Reserve boundaries. All pastoralist settlements within 2 km of the Reserve boundary (dotted li ne) are noted. (c) The location of the Talek region in the Maasai Mara National Reserve (dashed lines) and the community conservancies north of the Reserve (dark shaded regions). 14 Figure 2.1 contÕd. The two 4-km transects along which we monitored herbivores at biweekly intervals in the current study are noted with solid lines. Meteorological data Daily precipitation was monitored from a weather station in Talek. The effect of rainfall on herbivore numbers in this ecosystem is complex (Runyoro et al. 1995; Ogutu et al. 2008; 2009); however, years of extre me droughts may negatively influence herbivore numbers and the activity of pastoralists grazing their livestock . We calculated a drought index for each year as the total amount of rainfall in mm subtracted from the average of all years, divided by the stan dard deviation. Thus, a drought index value less than zero in any given year indicates below -average rainfall during that year. We acquired mean monthly minimum and maximum temperatures from the Narok meteorological station (~72 km northeast of the Reserve ) for the years 1989 to 2011. No temperature data were available for 2004 or 2005. Number of pastoralist settlements The communities around the Reserve are predominantly comprised of Maasai pastoralists living in traditional circular, fenced villages, cal led ÒbomasÓ (Lamprey & Reid 2004; Kolowski & Holekamp 2006) . To investigate the rate of human population growth in Talek, we identified all bomas within 2 km of the Reserve boundary in 1991, 2000, 2004, and 2012, using aerial photography, field surveys, and satellite imagery (Boydston et al. 2003a; 2003b; Kolowski & Holekamp 2006) . 15 Livestock No livestock were seen grazing on the open plains within Reserve borders in the Talek region from 1988 -1991. To estimate the numbers of livestock grazing inside the Reserve in Talek in later years, starting in 2000 we regularly drove throughout the Talek region and counted all sheep, goats, and cattle within Reserve boundaries (Kolowski & Holekamp 2009) . Counts occurred up to twice daily between 2000 -2008, and were performed between 0500 -1000 h and 1600-2000 h to account for variation in daily livestock grazing behavior. Starting in May 2008, all counts occurred between 1600 -2000 h and were only performed up to twice per month with one count taking place in the first half of the month, and one count taking place in the second half. For longitudinal analyses, we only included counts that occurred between 1600 -2000 h. Goats and sheep were combined (called ÒshoatsÓ) due to difficulties distinguishing between them in the field. ÒLivestock Ó included the total numbers of shoats and cattle in all analyses. Tourism We counted tourist lodges and interviewed lodge managers at all tourist facilities within 2 km of the Reserve boundary in the Talek region to document ch anges between 1988 and 2013 with respect to the total number of lodges and beds available to tourists. Statistical analyses All statistical analyses were performed in a Bayesian framework and conducted in R v. with JAGS (R Core Team 2015) . We analyzed temporal trends in the drought 16 index, average monthly minimum and maximum temperatures, and the number of bomas, tourist lodges , and beds available to tourists along the edge of the Reserve in the Talek region using general linear models . The climatic variables were modeled with a normal distribution and an identity link , and the number of bomas, tourist lodges, and beds available to tourists were modeled with a Poisson distribution and a log link . We modeled the climatic variables of drought and temperature , and the number of tourist lodges and beds for tourists, as a function of year, and the number of bomas as a function of year and year 2. Counts of livestock were overdispersed , so we analyzed them usin g a negative binomial distribution and a logit link. We modeled the number of livestock seen grazing inside park boundaries as a function of year, year 2, and year 3; in all analyses, we included year 2 and year3 to investigate non -linear trends in these para meters. To account for seas onal variation in temperature and livestock grazing, we fit month as a random effect. We developed hierarchical models in a Bayesian framework to examine trends in herbivore abundance and diversity in the Talek region, and to tes t predictions of hypotheses forwarded to explain this variation. To examine trends over time, we first used data from 1989 -2013 (25 years; 1989 was the first complete year of herbivore data) to model the total number of resident and migrant herbivores coun ted, species richness, and evenness, as a function of year, with transect identity coded as a fixed effect. Evenness was modeled using a log -normal distribution with an identity link, but the total number of resident and migrant herbivores and species rich ness were zero -inflated and overdispersed . We therefore modeled them with a zero -inflated negative binomial distribution with a log -link following Zipkin et al. (2010) . In this parameterization, 17 we defined yi,t,k,j as the dependent variable i in year t during biweekly interval k on transect j. Biweekly intervals ranged from 1 -24, and corresponded to the two -week long period during that year when the count took place (e.g. , days 1 -15 in January = biweekly interval 1). The mean of yi,t,k,j is µi,t,k,j = !i,t,k,j ¥ zi,t,k,j , where zi,t,k,j is a Bernoulli distributed random variable that determines whether or not a count is included in the model and is a function of an intercept ( $ 0) and transect identity ( $1). When zi,t,k,j = 1, the counts were modeled with a negative binomial distribution. Month of the survey was included as a random eff ect to account for the annual movements of wildlife within the Reserve. Next, using data from 2000 -2013 (14 years) we ran a second analysis to test hypotheses that might explain variation in herbivore abundance and diversity over time. This subset of data was selected because it encompassed the period during which we systematically monitored livestock inside Reserve boundaries, and therefore had near complete datasets for all of our covariates. Using the same distributions as in the previous analysis, we mo deled total numbers of resident and migrant herbivores counted, species richness, and evenness, as a function of the ecological and anthropogenic influences of the drought index, the minimum and maximum temperatures, the number of tourist lodges in the Tal ek region, the number of pastoralist settlements outside Reserve boundaries, the number of livestock grazing inside the Reserve in the Talek region , and an interaction between this last variable and drought. For years in which no data were available for bo mas, we extracted the predicted numbers from our generalized linear models. Missing data for the minimum and maximum temperatures of 2004, 2005, and 2012 were defined as the temperatures in the year preceding each missing year. We tested for the collineari ty of covariates using 18 PearsonÕs correlation coefficient s. The number of bomas and lodges in the Talek region between 2000-2013 were highly correlated ( SpearmanÕs correlation coefficient r = 0.97); to investigate the effects of tourist lodges and human pop ulation growth in the Talek region, we averaged the two values and created a new covariate ( ÒdevelopmentÓ). There were no significant correlations between any other covariates ( SpearmanÕs correlation coefficient r < 0.7). We evaluated the effects on our de pendent variables of each independent variable during the 2 years prior to the count. All covariates were mean -standardized by 2-year average to examine the relative impact of each independent variable on our dependent variables. Thus, we modeled the estim ated mean herbivore count or diversity measure per sample ( !i,t,k,j ) with the function: !i,t,k,j = % 0[month] i + %1i ¥ drought t-2 + %2i ¥ tmin t-2 + %3i ¥ tmax t-2 + % 4i ¥ developmentt-2 + %5i ¥ livestock t-2 + % 6i ¥ (livestock t-2 ¥ drought t-2) + % 7i ¥ trans ect j where ! i,t,k,j is a function of the intercept ( % 0) and ecological and anthropogenic covariates ( % 1 through % 7). Each model was fit separately by MCMC. We used uninformative priors for all of our covariates, and each model was run with 3 chains, for 100,000 iterations, with a 50,000 burn -in. Chains were thinned by 10, and ! values were examined to ensure model fit (Gelman & Hill 2007). Posterior parameter estimates with 95 % credible intervals (CIs) that did not overlap zero were considered to be statistically significant. In 19 all analyses, year was coded as a time -series variable so that each year indicated an increase of on e time step. Results Herbivore abundance and diversity From January 1989 through December 2013, we sampled the abundance, diversity, and evenness of mammalian herbivores in Talek biweekly, a total of 1104 times (mean ± standard error replicates per transe ct = 552.0 ± 6). Species richness and evenness values were available for 902 and 773 of these samples, respectively. Counts of herbivore abundance and diversity varied among months, with a distinctive trough in the number of resident herbivores counted during months when migrant herbivores were present (Figure 2.2). 20 Figure 2.2. Temporal trends in the modeled count s of (a) resident and (b) migrant herbivores, (c) species richness, and (d) the Shannon index in the Talek region by month between 1989 -2013. Mean ± 95 % credible intervals are displayed. We found evidence for significant declines in the total number of resident and migrant herbivores counted, species richness, and evenness in Talek over time (Table 2.1, Figure 2.3). The numbers of resident and migrant herbivores counted and species richness varied with transect identity ( Table 2.1, Figure 2.3); the transect located in long-grass habitat had significantly fewer resident herbivores and lower species 0100200300400Month a.123456789101112Resident herbivores counted 0100200300400Month b.123456789101112Migrant herbivores counted 3.03.54.04.55.0Month c.123456789101112Species richness 0.40.50.60.70.80.9Month d.123456789101112Shannon index 21 diversity, but a larger number of migrant herbi vores, than did the transect in short -grass habitat (Table 2.1, Figure 2.3). Species evenness did not vary with transect identity (Table 2.1). Transect identity had a significant effect on the inclusion parameter zi,t,k,j for all dependent variables, indi cating that species richness and herbivore numbers were more likely to be greater than zero on the transect in short -grass habitat ( expected mean posterior [95 % CI] residents: intercept = 6.78 [4.65, 9.6], transect = -5.69 [-8.5, -3.54]; migrants: interce pt = 4.22 [1.94, 8.35], transect = -3.47 [-7.48, -1.38]; species richness: intercept = 6.93 [4.7, 10.66], transect = -5.69 [-9.4, -3.44]). Table 2.1. Mean posterior parameter estimates ( and 95 % credible intervals ) of random and fixed effects from the hie rarchical models that investigated trends in the numbers of resident and migrant herbivores, richness, and evenness from 1989 -2013 in the Talek region of the Maasai Mara National Reserve. Significant posterior estimates , not including the intercept, are in dicated in bold font (95 % credible intervals not overlapping zero). Month of the year was modeled as a random effect on the intercept ; here we present the average for all 12 months . Parameter estimates are for the transect located in long-grass habitat. Parameter Residents Migrants Richness Shannon Intercept 5.13 (4.87, 5.39) 3.45 (2.91, 4.06) 1.43 (1.33, 1.52) -0.48 (-0.63, -0.34) Trend -0.01 (-0.02, 0) -0.05 (-0.07, -0.02) -0.02 (-0.02, -0.01) -0.01 (-0.02, 0) Transect -1.33 (-1.49, -1.16) 0.56 (0.22 , 0.91) -0.34 (-0.42, -0.26) 0 (-0.1, 0.1) 22 Figure 2.3. Temporal trends in the mean expected yearly count of (a) resident and (b) migrant herbivores, (c) species richness, and (d) the Shannon index in the Talek region between 1989-2013. The tran sect in short grass habitat is indicat ed with solid lines and the transect in long grass habitat is indicat ed with dashed lines (a-c). We present the average for both transects because there was no statistical difference in the Shannon index between transe cts (d) . All declining trends were significant (95 % credible intervals did not overlap zero for the parameter of trend over time ). 050100150200250Year 1989199319972001200520092013Resident herbivores counted a.050100150200250Year 1989199319972001200520092013Migrant herbivores counted b.0123456Year 1989199319972001200520092013Species richness c.0.500.550.600.650.700.75Year 1989199319972001200520092013Shannon index d. 23 Meteorological data Drought index values < 0 were recorded for nearly half of the study years, but did not increase in frequ ency over time ( Table 2.2, Figure 2.4). Both yearly mean minimum and maximum temperatures recorded in Narok between 1989 -2011 increased over time (Table 2.2, Figure 2.4). Mean minimum temperatures increased by 2.6 ¼C and mean maximum temperatures increased by 1.4 ¼C. 24 Table 2.2. Mean posterior parameter estimates and 95 % Credible Interval s (95 % CI) of the models investigating the temporal trends in ecological and anthropogenic factors in the Talek region. A negative binomial distribution with a logit lin k was used for livestock, and a Poisson distribution with a log link was used for the number of bomas, tourist lodges, and beds for tourists; all other analyses were performed using a normal distribution and an identity link . All posterior parameter estima tes are presented on the normal scale. The parameter Year indicates the amount of change in each subsequent year. Significant posterior parameters are indicated in bold font (95 % CI not overlapping 0 ). Parameter Estimat e 95 % CI Drought Intercept 0 (-0.428, 0.42) Year 0.024 (-0.035, 0.082) Max. temperature Intercept 24.832 (24.39, 25.276) Year 0.032 (0.013, 0.051) Min. temperature Intercept 10.272 (9.786, 10.755) Year 0.074 (0.054, 0.095) Bomas Intercept 60.039 (49.949, 71.593) Year 4.036 (2.399, 6.292) Year 2 0.064 (-0.002, 0.003) Livestock Intercept 512.816 (418.287, 632.017) Year 18.646 (-0.8, 50.419) Year 2 -20.949 (-19.643, -23.643) Year 3 2.049 (1.595, 3.311) Tourist l odges Intercept 5.641 (4.735, 6.659) Year 0.409 (0.238 , 0.649) Beds for tourists Intercept 468.717 (460.356, 477.231) Year 26.502 (25.057, 28.498) 25 ! Figure 2.4. Yearly temporal trends in (a) the drought index in the Talek region of the Maasai Mara National Reserve, and (b) mean ± standard error ( SE) maximum and (c) minimum temperatures in Narok, Kenya. Drought index 1988199019921994199619982000200220042006200820102012-3-2-101231988199019921994199619982000200220042006200820102012222324252627Mean ± SE yearly maximum temperature (¼C) b.Year 1988199019921994199619982000200220042006200820102012789101112Mean ± SE yearly minimum temperature (¼C) c. 26 Figure 2.4 contÕd. Solid lines are the polynomial equations fit to the data using general linear and generalized linear mixed -effects models in R , and dashed lines indicate the 95 % credible interval a round this equation .The drought index is the number of standard deviations from the mean precipitation for the entire study where a given yearÕs precipitation was observed to fall . Number of pastoralist settlements The human population along the Reserve b oundary increased from 1991 to 2012 (Table 2.2, Figures 2.1 & 2.4). This represents growth in the local Maasai community by over 406 % during the course of this study. Livestock The average number of livestock counted grazing daily inside the Reserve in the Talek region increased from 0 in 1988, to 2,218.85 in 2013 ( Table 2.2, Figure 2.5). Furthermore, the number of shoats seen grazing inside park boundaries increased over time. Whereas cattle represented 85.0 % of all livestock counted in 2000, they repre sented only 66.7 % in 2013. 27 ! Figure 2.5. Temporal trends in the (a) total number of pastoralist settlements along the edge of, and (b) mean numbers (± standard error ) of livestock grazing daily inside, Reserve boundaries in the Talek region between 1988 -2013. Count of bomas 198819911994199720002003200620092012050100150a.Year Average number of livestock 19881991199419972000200320062009201201000200030004000b. 28 Figure 2.5 contÕd. Solid lines represent the polynomial equations fit to the data using general linear and generalized linear mixed -effects models in R , and dashed lines indicate the 95 % credible interval around this equation. Tourism The number o f tourist facilities and beds available for tourists increased between 1988 and 2013 (Table 2.2, Figure 2.6), representing increase s of 500 % and 366.4 %, respectively. These increases started in 1997 and continue to the present (Figure 2.6). 29 ! Figure 2.6. Temporal trends in the number of tourist lodges ( ") and beds available to tourists ( o) in the Talek region of the Maasai Mara National Reserve, Kenya, between 1988-2013. Solid lines represent the polynomial equations fit to the data using general linear models in R , and dashed lines indicate the 95 % credible interval around this equation. Year # of Lodges ( !)1988199119941997200020032006200920120510152002004006008001000# of Beds (o) 30 Hypothesis testing We modeled variation in the abundance, diversity, and evenness of mammalian herbivores in the Talek region between 2000 and 2013 as a function of th e ecological and anthropogenic factors monitored there during the 2 years before each herbivore count. In this subset of data, transect identity significantly predicted species richness and the number of resident herbivores on transects, but not any other measure; species richness was higher, and there were more resident herbivores counted, on the transect in short -grass habitat than on the transect in long -grass habitat (Table 2.3). Transect identity also had a significant effect on the inclusion parameter zi,t,k,j for the abundance of resident herbivores and species richness, but not for the number of migrant herbivores (expected mean posterior [95 % CI] residents: intercept = 5.79 [3.60, 9.91], transect = -4.83 [-8.94, -2.62]; species richness: intercept = 5.89 [3.80, 5.64], transect = -4.58, [-8.03, -2.47]; migrants: intercept = 4.25 [1.50, 8.46], transect = -2.92 [-7.17, 0.64]). Although none of our ecological or anthropogenic factors predicted variation in the numbers of migrant herbivores or species e venness, many factors were important in predicting resident herbivore abundance and species richness. The habitual grazing of livestock inside the Reserve and increased rainfall had significant negative effects on both the abundance of resident herbivores and species richness ( Table 2.3), yet the interaction between these two factors had no effect ( Table 2.3). Increases in minimum and maximum temperatures also positively affected species richness and the numbers of resident herbivores on transects ( Table 2.3). Surprisingly, t he documented increase in development of tourist facilities and the number of pastoralist settlements along the 31 edge of the Talek region had no effect on any of our dependent variables. Of our independent variables, livestock grazing had the largest, significant negative effect on the abundance of resident herbivores and species richness (Figure 2.7). Table 2.3. Mean posterior parameter estimates (and 95 % credible intervals ) from the hierarchical models that tested hypotheses to explain the variation in herbivore diversity and abundance in the Talek region. Statistically significant parameter estimates are highlighted in bold font (95 % credible intervals that did not overlap zero). Month of the year was modeled as a random effect on the intercept ; here we present the average for all 12 months. Parameter estimates are for the transect located in long -grass habitat. Parameter Residents Migrants Richness Shannon Intercept 4.73 (4.33, 5.15) 2.84 (1.88, 3.94) 1.26 (1.12, 1.4) -0.55 (-0.71, -0.39) Drought -0.93 (-1.26, -0.6) 0.16 (-0.63, 0.94) -0.28 (-0.44, -0.13) -0.03 (-0.21, 0.15) Development -0.1 (-0.4, 0.2) 0.36 (-0.41, 1.09) -0.05 (-0.19, 0.09) 0.03 (-0.14, 0.2) Minimum temperature 0.98 (0.22, 1.76) -0.18 (-1.87, 1.59) 0.33 (0.02, 0.6 4) 0.11 (-0.28, 0.48) Maximum temperature 0.81 (0.13, 1.52) -0.85 (-2.42, 0.76) 0.32 (0.02, 0.63) 0.16 (-0.2, 0.51) Livestock -1.36 (-2.47, -0.31) 0.94 (-1.55, 3.39) -0.53 (-0.99, -0.06) -0.31 (-0.85, 0.24) Livestock x Drought -0.38 (-0.83, 0.07) -0.1 (-1.15, 0.93) -0.14 (-0.34, 0.07) 0.03 (-0.21, 0.26) Transect -1.34 (-1.6, -1.08) -0.18 (-0.86, 0.48) -0.37 (-0.49, -0.25) 0.05 (-0.09, 0.18) 32 Figure 2.7. The relative importance of the ecological and anthropogenic variables affecting the abunda nce of resident and migrant herbivores, species richness, and Shannon index. Mean posterior estimates and 95 % Credible Intervals (CIs) are shown. Discussion This study presents new and important data d ocument ing trends in overall mammalian herbivore dive rsity in the Talek region of the Maasai Mara National Reserve. Species richness, species evenness, and the numbers of resident and migrant herbivores declined significantly during the course of this study. Although collected completely independently and us ing different methods from those of earlier studies, our Mean parameter estimate ± 95 % CI Drought Minimum Maximum Development Livestock Livestock x Transect temperature temperature Drought Factor -3-2-10123Resident herbivores Migrant herbivores Species richness Shannon index 33 data are consistent with results from previous research on temporal changes in herbivore numbers in the Reserve (Ottichilo et al. 2000; Ogutu et al. 2005; Norton -Griffiths et al. 2008; Ogutu et al. 2009; 2011) . In the current study, we tested specific hypotheses that might explain changes in herbivore diversity and abundance in the Tale k region. Although none of our ecological or anthropogenic factors predicted the numbers of migrant herbivores or species evenness, many of our covariates were important in predicting the abundance of resident herbivores and species richness . Increases in minimum and maximum temperatures positively affected the abundance of resident herbivores and species richness. The Talek region is primarily covered with C 4 grasses (McNaughton 1983) , and higher local temperatures are like ly to increase their nutrient concentrations over time (Ritchie 2008) . Thus, in the short -term, increased local temperatures may provide a small benefit to wild herbivores. Species richness and the number of resident herbivores in our study were negatively affected by rainfall. The relationship between rainfall and herbivore abundance in the Reserve is comple x, and it varies among species and herbivore age classes (Ogutu et al. 2008; 2009) . Our results add to these previous findings, but are not straightforward . We do not yet know the mechanism for the negative effect of rainfall in the current study , but possible explanations may include the susceptibility of wildlife to disease during wet years , or a negative interaction between localized habitat degradation and extremely wet years. Further research is needed to shed more light on this . The number of livestock grazing in the Reserve had the strongest negative effect on species richness and on nu mbers of resident herbivores in Talek (Figure 2.7). 34 Although illegal since the Reserve was established (Talbot & Olindo 1990) , livestock grazing has always been a part of this ecosystem (Reid 2012) . Small numbers of livestock were present inside the Reserv e in Talek before 1992, but herders grazed them surreptitiously in thickets along the northern border . Thus, livestock grazing inside the park went from being low -intensity and restricted in distribution on the open plains of Talek from 1988 -1991, to an average of more than 2,000 livestock counted in open plain regions daily twenty years later in 2013. One recent livestock count in early 2015 revealed 20,800 livestock grazing inside the Reserve in the Talek area in a single day, so livestock grazing within Reserve borders still appears to be increasing. Identifying the exact mechanism for the negative effect of livestock on wildlife populations in the Reserve will demand further research, but might stem from competition between livestock and wildlife for nut rients (Prins 1992; Veblen 2008; Riginos et al. 2012), the compaction of soils in heavily grazed regions (Figure 2.8), and/or distinctive changes in vegetation and habitat due to overgrazing (Homewood & Rodgers 1988; Kiage 2013) . Regardless of the specific mechanism, livestock grazing most likely affects important food resources and other habitat features salient for wild ungulates. !! 35 a. b. Figure 2.8. The intensive grazing taking place in the Talek region. (a) Multiple herds like this graze inside the Reserve daily. (b) The effects of intensive, sustained grazing in the Talek region on vegetation and soils, as seen from a satellite orbiting the earth in space (Photo credit: Google Earth). 36 Figure 2.8 contÕd. The near-vertical swaths across this landscape are the tracks left by large cattle herds moving back and forth into the Reserve from bomas situated outside the norther n Reserve boundar y, indicated here by the jagged line of riverine vegetation . Although our analyses suggest the number of tourist lodges and pastoralist settlements did not significantly predict declines of resident and migratory herbivore populations and diversity in Talek, habitat change and degradation north of the Reserve were believed to contribute to wildlife declines in earlier studies (Ottichilo e t al. 2000; Serneels et al. 2001; Serneels & Lambin 2001; Ogutu & Dublin 2002; Ogutu et al. 2009; 2011). Many resident herbivores regularly use pastures outside of the Reserve due to displacement by migratory animals, to gain access to better forage, or f or wet season dispersal ranges (Stelfox et al. 1986; Bhola et al. 2012a; 2012b) . The rapid habitat changes due to tourism and human population growth i n and around Talek, including fencing, may ultimately restrict important wildlife corridors , and prevent these seasonal movements from occurring. Furthermore, there are likely to be other negative indirect effects from the increase in pastoralist settlemen ts and tourism infrastructure along the Talek River. The Talek River is the northern boundary of the Reserve in the Talek region, and the only semi -permanent watercourse in this area (Figure s 2.1 and 2.8). Habitat changes associated with an increase in the numbers of pastoralist settlements and tourist lodges along its banks may ultimately trigger broad ecological changes due, for example, to declining water quality. Future research should investigate the indirect effects this habitat change might have on w ildlife. 37 The results from the current study are representative of what is occurring in the Talek region, but not throughout the Reserve as a whole. The Talek and Sekenani regions have rapidly expanding human populations along Reserve boundaries (Reid et al. 2003; Lamprey & Reid 2004) . However, other parts of t he greater Mara ecosystem remain pristine because they are much less strongly affected by people, including the entire area of the Reserve lying west of the Mara River, which is managed by a nonprofit organization called the Mara Conservancy (Walpole & Leader -Williams 2001) . When the Mara Conservancy initially took over management of the western portion of the Reserve in 2001, the environment was degraded due to lax control of livestock grazing and poaching of wild ungulates (Ogutu et al. 2009) . After initiating strict limits on grazing within Reserve boundarie s and increasing anti -poaching patrols, this portion of the Reserve is now remarkably unaffected by anthropogenic disturbances aside from visitation by tourists. Areas north of the Reserve made similar changes with the establishment of several private cons ervancies starting in 2005 (Figure 2.1). The merging of these rangelands has transformed ~ 938 km2 of area once open for livestock grazing into heavily managed wildlife preserves (Figure 2.1). The long -term effects of these conservancies on wildlife within the Reserve have yet to be elucidated. However, it is possible that the development of these restricted -grazing regions might actually be causing Maasai pastoralists to graze their livestock within the Reserve more frequently than they did before es tablishment of these conservancies . Local wildlife management authorities have no control over abiotic factors such as rainfall or temperature; however, they can and should control the frequency and intensity of livestock grazing within Reserve boundaries. Prio r to the creation of the 38 Reserve, livestock were an integral part of the Mara ecosystem (Ho mewood & Rodgers 1991; Reid 2012), and they are unlikely to disappear from it any time soon. At decreased levels, there is a possibility for mutual benefits from shared grazing by livestock and native ungulates (Prins 2000; Arsenault & Owen -Smith 2002; Riginos et al. 2012). Livestock grazing at moderate levels and rotational grazing patterns can enrich savanna ecosystems through habitat modification (Adler et al. 2001; Augustine et al. 2011). Yet, without strict management, enforcement, and compliance, habitual and intensive grazing of livestock has the potential to competitively supp ress wild herbivore populations (Prins 1992; 2000; Butt & Turner 2012) , and can lead to long -term declines in the abundance and diversity of native wildlife (Reid 2012; Riginos et al. 2012) . There are many important and unanswered questions emerging from the current research that future study of variation in management regimes could potentially address. These include understan ding 1) what effects the private conservancies are having on the intensity of livestock grazing within Reserve boundaries, 2) whether wildlife numbers and diversity will recover if livestock grazing is controlled in the Reserve, and 3) what long-term effec ts the development of tourism infrastructure are having on wildlife populations and ecosystem function. In other regions of Kenya, maintenance of the traditional nomadic lifestyle of pastoralists helps mitigate conflicts between livestock and carnivores (Schuett e et al. 2013b), and reduces the probability of long -term changes in herbivore numbers and diversity. Research is desperately needed in the Mara ecosystem to understand how traditional Maasai pastoralists and their livestock can best coexist with native w ildlife, and how tourism practices can work to have a minimal impact on resident wildlife. However, without the rapid collection of data to address 39 these questions and the enactment of management reforms that limit or alleviate some of the grazing pressure and tourism development in Talek, numbers and diversity of wild ungulates are likely to continue to decline in the eastern Maasai Mara. 40 CHAPTER 3: Effects of anthropogenic disturbance on long -term trends in the demography and behavior of lions and spot ted hyenas in the Maasai Mara National Reserve, Kenya 41 Introduction The conservation outlook is currently grim for large carnivores throughout the world. Their populations have been declining in most ecosystems as a result of habitat change, direct confli cts over livestock, utilization of their body parts for subsistence and trade, and depletion of their prey (Ripple et al. 2014) . Although there have been many recent examples of the potential for coexistence between carnivores and people in human-dominated landscapes (Carter et al. 2012; Schuette et al. 2013b; Yirga et al. 2013; Dorresteijn et al. 2014) , and although some carnivore populations may actually be rebounding in the face of growing human populations (Chapron et al. 2014) , most carnivore species known to be critical for maintaining ecosystem function have undergone major declines or range contractions in r ecent decades (Ripple et al. 2014) . The reduction or complete extirpation of large carnivores can have ecosystem -wide consequences due to the lessening or removal of the top -down control many of them exert over species at lower trophic levels in natural ecosystems (Estes et al. 2011; Ripple et al. 2014). One common example of the restructuring of natural ecosystems that occurs when apex predators decline is a specific t ype of trophic cascade called a mesopredator release (Cr ooks & Soul” 1999) ; this involves an increase in numbers of smaller carnivores that are ordinarily limited by the apex predators. In addition to mesopredator releases, the removal of large carnivores can trigger predator -mediated trophic cascades in other sympatric wildlife populations (Hebblewhite et al. 2005; Johnson et al. 2007; Berger et al. 2008) . The effects of these trophic cascades have been well documented in many parts of the world, and can include declines in bird 42 populations, altered vegetation structure, and potentially even climate change (Estes & Palmisano 1974; Beschta & Ripple 2009; Wilmers et al. 2012) . In many developing nations, Protected Areas (PAs) play a crucial role in the conservation of large carnivores (Woodroffe & Ginsberg 1998; Packer et al. 2013) . However, rapid human population growth around PAs has the capacity to intensi fy anthropogenic threats to wildlife within reserve boundaries (Wittemyer et al. 2008; Newmark 2008; Craigie et al. 2010) . Large carnivore populations are extremely vulnerable to these types of threats due to the ir expansive home ranges, slow life histories, and large energy demands (Ripple et al. 2014; Woodroffe and Ginsberg 1998). Thus, it is critical to understand how anthropogenic disturbances around existing PAs might be affecting carnivore populations within PA borders. The Mara -Serengeti ecosystem is a stronghold for the conservation of large carnivores in east Africa (Ogutu & Dublin 2002; Riggio et al. 2013) , and revenues from tourists hoping to see large carni vores there represent a key source of foreign exchange for local and national economies (Karanja 2003; Norton -Griffiths et al. 2008; Polasky et al. 2008). The Maasai Mara National Reserve (henceforth, the Reserve) comprises the northernmost portion of the Mara -Serengeti ecosystem, and has historically contained some of the highes t densities of large carnivores in the world due to abundant prey throughout the year (Ogutu & Dublin 2002; Craft et al. 2015) . However, activity associated with recent human population growth outside of the Reserve has been affecting wildlife populations within Reserve boundaries. Starting in the 19 60s, habitat change and l ivestock grazing inside and outside of the Reserve have been implicated in the widespread declines in herbivore populations within the Reserve (Ottichilo et al. 43 2000; Ogutu et al. 2009; 2011; see also Chapter 2 of this dissertation) , and have been closely associated with behavioral (Boydston et al. 2003a; Kolowski et al. 2007; Kolowski & Holekamp 2009) , physiological (Van Meter et al. 2009) , and demographic changes in spotted hyenas ( Crocuta crocuta ; Watts & Holekamp 2009; Pangle & Holekamp 2010) , the most abundant large predator in this ecosys tem. Limited information is available on how anthropogenic disturbance may be affecting other large carnivores in this ecosystem, such as lions (Panthera leo; but see Ogutu et al. 2005), and this is an important ar ea for research. The Reserve is a good place in which to study effects of anthropogenic disturbance on large carnivore populations within PAs due to its high densities of lions and spotted hyenas and the rapid human population growth around its borders (Ogutu & Dublin 2002; Lamprey & Reid 2004; Ogutu et al. 2005; Watts & Holekamp 2009) . Throughout their range, lions and hyenas exhibit a high degree of dieta ry and spatial overlap (Hayward & Kerley 2005; Hayward 2006; P”riquet et al . 2014), resulting in intense interference and exploitation competition and intraguild predation (Kruuk 1972; Cooper 1991; Watts & Holekamp 2008; 2009; P”riq uet et al. 2014). Lions have historically represented the greatest natural source of mortality to spotted hyenas (Kruuk 1972; Watts & Holekamp 2008; 2009; P”riquet et al. 2014) . Conflict between lions and people can result in declines in lion population numbers (Woodroffe 2001; Woodroffe & Frank 2005) , and in areas where lion numbers have declined significantly, spotted hyenas can more easily maintain possession of their kills and kleptoparasitize food from lions (Watts & Holekamp 2008) . This in turn enhances their reproductive success by permitting younger weaning ages of cubs and shorter inter -birth intervals 44 (Watts & Holeka mp 2008) . This suggests that the short -term effects of people on populations of large carnivores in the Reserve may be negative for lions, but might indirectly be positive for hyenas. In addition to changing population sizes, one other way in which lions and spotted hyenas may respond to anthropogenic disturbance in the Reserve is through their space -use. Although it is affected by many factors, including patterns of rainfall, prey availability and moonlight (e.g., Stratford & Stratford 2011; Cozzi et al. 2012; Oriol -Cotterill et al. 2015) , the space -use of large carnivores can also be affected by anthropogenic disturbance. For example, large carnivores in areas of high anthropogenic activity may shift their activity patterns to be more nocturnal to avoid conflict with humans (e.g., Van Dyke et al. 1986; Frank & Woodroffe 2001; Carter et al. 2012; Rasmussen & Macdonald 2012) , or alter their space -use to completely avoid areas of intense anthro pogenic activity (e.g., Van Dyke et al. 1986; Schuette et al. 2013b; 2013a). Previous research in the Re serve has shown spotted hyenas modify their behavior in areas of intensive livestock grazing. Spotted hyenas whose group territory was used daily for livestock grazing were more likely to be active at night than during day -light hours (Boydston et al. 2003a; Kolowski et al. 2007) , and were found more often in or near dense vegetation than were cons pecifics whose territory was not grazed by livestock (Boydston et al. 2003a; Kolowski & Holekamp 2009) . Unfortunately, no research exists on how lions may be affected by increases in livestock grazing in the Reserve, but previous research has shown that lions shift their temporal patterns of activity to avoid encountering people (Schuette et al. 2013b; Oriol -Cotterill et al. 2015) . Lions and hyenas may respond to the effects of anthropogenic disturbanc es within and 45 around the Reserve by moving farther away from its boundaries, effectively diminishing the size of the PA for them. As the numbers of livestock grazing in the Reserve continue to increase (Boydston et al. 2003a; Kolowski & Holekamp 2009 ; Chapter 2 of this dissertation) , and habitat change outside of the Reserve accelerates (Lamprey & Reid 2004; Norton -Griffiths et al. 2008) , it is critical to understand how lions and hyenas are affected by disturbances along Reserve boundaries. Our goals were to investigate trends in the population sizes of spotted hyenas and the number of lions sighted within the Reserve, determine how these trends varied in areas characterized by varying amount of anthropogenic disturbance, and examine how the space -use of these large carnivores inside the Reserve is affected by anthropogenic disturbance inside and outside Reserve boundaries. To do this we used data collected since 1988, and exploited naturally -occurring variation in the exposure of these large carnivores to anthropogenic disturbance within Reserve boundaries. The effects on carnivores inside the Reserve of anthropogenic activity associated with human populations living immediately outside Reserve borders are expected to increase with the carnivoresÕ proximity to Reserve boundaries. Therefore the distances at which lions and hyenas are found from Reserve borders can be used as a proxy to assess their respective sensitivities to anthropogenic disturbance. Because spotted hyenas exhibit great behavioral flexibility (Holekamp & Dloniak 2010) , we predicted that their numbers would be less strongly affected by anthropogenic disturbance than would numbers of lions. Given that lions exhibit less flexible behavior, have a larger body size, and have slower life histories, we anticipated their numbers might decline in disturbed areas (Woodroffe 2001; G onz⁄lez -Su⁄rez & Revilla 2013) . Over time we expected this 46 would result in declines in the population sizes of lions in areas of comparatively high anthropogenic disturbance. We also predicted that, in areas of more intensive anthropogenic activity, spott ed hyenas and lions would both be more likely to be found deeper in the Reserve and thus farther away from people. Here we present, to our knowledge, the first study documenting how anthropogenic disturbances along the edge of a PA can influence the number s and space -use of sympatric lions and hyenas concurrently, and how the differential sensitivities of these large carnivores to this disturbance may be manifested. Methods Study site The Reserve is primarily comprised of open grassland interspersed with r iparian areas. It has traditionally supported large herds of resident and seasonally migratory herbivores, together with populations of small and large carnivores that are present year-round (Bell 1971; Sinclair & Norton -Griffiths 1979; Stelfox et al. 1986; Craft et al. 2015). Rainfall patterns in the Reserve are bimodal, with most rain falling in November -December and March -May. There is considerable naturally -occurr ing variation in the exposure of carnivores to anthropogenic disturbance within park boundaries stemming from differences in management regimes and proximity to the edge of the Reserve. The eastern side of the Reserve is managed by the Narok County Governm ent. Within this portion of the Reserve ( Figure 3 .1), the Talek region in particular is an area near the Reserve border that in recent decades has undergone enormous change involving increases in the 47 number of human settlements and tourist facilities outsi de the Reserve, as well as an exponential increase in livestock grazing inside Reserve boundaries (Boydston et a l. 2003a; Lamprey & Reid 2004; Kolowski & Holekamp 2009 ; Chapter 2 of this dissertation) . In contrast to the Talek area, other portions of the Reserve are far less affected by anthropogenic activity. In particular, the portion of the Reserve west of the Mara River, called Òthe Mara Triangle,Ó is remarkably pristine (Figure 3.1) . In 2001, management of The Mara Triangle was assigned to the Mara Conservancy, and the wildlife in this part of the Reserve experience virtually no anthropogenic activity at all aside from visitation by tour vehicles. Here we were able to exploit these differences in anthropogenic activity to test hypotheses suggesting how large carnivores within Reserve boundaries are being affected by anthropogenic disturbance . 48 ! Figure 3.1. Sit es of carnivore monitoring in the Maasai Mara National Reserve, Kenya. The approximate boundaries of the six territories defended by members of our study hyena clans are indicated with dashed lines. Study clans in the Mara Triangle (shaded region), which i s managed by the Mara Conservancy, are Serena North (SN), Serena South (SS), and Happy Zebra (HZ). Clans in regions managed by the Narok County Government (no shading) are Talek West (TW), Fig Tree (FT), and Mara River (MR). The TW clan is located in the T alek region of the Reserve. The star (!) indicates the closest town bordering the Talek region (the town of Talek). 49 Data collection Demography and behavior of spotted hyenas Spotted hyenas live throughout sub -Saharan Africa in social groups, called clans, containing an average of 2 9 individuals (Holekamp & Dloniak 2010) . Clans are comprised of adult females and their juvenile offspring, and several immigrant males. Males usually disperse after reaching sexual maturity, which occurs at approximately 24 months of age (Smale et al. 1997; van Horn et al. 2003; Boydston et al. 2005) , but females are philopatric except in cases of rare clan fission events, when groups of females split off from the parent clan to form entirely new clans (Holekamp et al. 1993; Frank et al. 1995; Hofer & East 2003; Boydston et al. 2005; Hıner et al. 2007) . Each clan is structured by a linear matrilineal dominance hierarchy in which natal animals dominate immigrant males; immigrant males hold ranks based on their tenure in the clan (Smale et al. 1997; East & Hofer 2001) . Clans of spotted hyenas have been monitored in the Maasai Mara National Reserve as part of a longitudinal study that began in 1979 (Frank 1986) . Here we monitored six different clans between 1988 and 2013 throughout the Reserve, in areas characterized b y varying levels of anthropogenic activity. We studied five clans exposed to low levels of anthropogenic disturbance (Serena North, Serena South, Happy Zebra, Mara River, Fig Tree), and one clan exposed to a significantly higher level in the Talek region (Talek West ; Figure 3 .1). Wild animals in the Talek region are exposed to thousands of livestock grazing daily inside the Reserve (Boydston et al. 2003a; Kolowski & Holekamp 2009 ; Chapter 2 of this dissertation) , and because they live near 50 the edge of the Reserve , they are experiencing rapid human population growth and rapid development of tourism infrastructure (Chapter 2 of this dissertation) . We observed hyenas daily in each clan during morning (0500 -1000 h) and evening (1600 -2100 h) observation periods. During these two periods of observation, we drove 1-2 vehicles systematically throughout the territor ies of one or more study clans, recording the GPS locations of all hyenas enco untered. In 2013, size of the clan territories monitored ranged from 28.29 to 77.04 km 2 (Figure 3 .1). All hyenas in each study clan were individually recognized based on their unique spot patterns, scarring, and ear damage. When a natal animal was seen for the first time, its birthdate was estimated to ± 7 days (Holekamp et al. 1996) , and it was sexed based on the morphology of its erect phallus (Frank et al. 1990) . Immigrant males were assigned clan membership after being present for at least 6 months. With our intensive, year -round monitoring, individuals were considered dead or missing if they were not seen for six continuous months. Their date last seen then became their death or disappearance date. To assess demographic change throughout the study, mean monthly clan size was calculated as the total number of individuals in all age classes present during each month in which a particular clan was observed. In addition to non-invasively monitoring clan demography and the locations of hyenas via daily observations, starting in 1991 multiple individuals were immobilized with Telazol (6.5 mg/kg) administered in a plastic dart fired from a CO 2 powered rifle (Telinject Inc., Saugus, California), and fitted with VHF (Telonics Inc., Mesa, AZ, USA) or GPS radio collars with VHF functionality (Vectronic Aerospace, Berlin, Germany). We utilized radio telemetry to assist in locating hyenas in the Talek region, and recorded the 51 location, composition, and group sizes of hyenas when individuals were located. A visual confirmation was not always possible, and in these cases, the location was triangulated and localiz ed to within 200 m2. Estimati ng lion abundance and documenting their use of space We used two different metrics to investigate trends in lion abundance in the Reserve because records were not maintained on individual lions as they were for spotted hyenas . First, between 1988 and 2013, we monitored the number of times lions were found < 200 m from spotted hyenas each month within the boundaries of the territory defended by the Talek West clan of spotted hyenas. Second, we also recorded the GPS locations of all lions seen during morning and evening observation periods described above within the territories of four of our hyena study clans regardless of whether or not spotted hyenas were also present with the lions. This additional sampling occurred between 2004 -2013 in the territory defen ded by the Talek West clan of spotted hyenas, and between 2009 -2013 in the territories defended by the Serena North, Serena South, and Happy Zebra clans of spotted hyenas. Monitoring of ecological and anthropogenic variables To distinguish between anthrop ogenic and natural ecological influences on the distances from the Reserve boundary at which lions and hyenas were sighted, in the Talek region we calculated the mean number of herbivore prey present in the territory of the Talek West clan every month, the amount of rainfall, and the mean number of livestock grazing in the territory of the Talek west clan. Three 4 -km line -transects were 52 sampled biweekly in Talek on which all mammalian herbivores were counted within 100 m of each transect centerline. Further information on our methods can be found in previous publications (e.g., Holekamp et al. 1999; Boydston et al. 2003a; Van Meter et al. 2009). We calculated the monthly average of the total numbers of herbivores counted in Talek on all three transects. In each month, we also calculated the total amount of rain that fell at a weather station in the Talek region. Starting in 2000, we performed livestock counts to estimate the total number of livestock grazing in the Talek region o f the Reserve. Observers systematically drove throughout the Talek West territory and recorded the numbers of cattle, sheep, and goats grazing within park boundaries. Counts occurred up to twice daily between 2000 -2008, and were performed between 0500-1000 h and 1600-2000 h. Starting in May 2008, all counts occurred between 1600-2000 h and were only performed up to twice per month , with one count taking place in the first half of the month and one count taking place in the second half. No livestock were pre sent in the territory of any other study clan. We mean -standardized the monthly estimates of herbivore abundance, the amount of rainfall, and the number of livestock recorded in Talek for use in subsequent analyses. Statistical analyses All analyses were conducted with generalized linear and generalized linear mixed-effects models (see parameterizations below) . We fit all of our models using MCMC analyses in R with JAGS (R Core Team 2015) . For all anal yses, we used uninformative priors, ran our models with three chains for 20,000 iterations with a 10,000 burn -in, and thinned by 10. We examined !!values for all parameters to ensure 53 model fit (Gelman & Hill 2007) . For all analyses, we present posterior parameter estimates and credible intervals (CI s) on the normal scale . Parameter estimates for which the 95 % CI did not overlap zero were considered to be statistically significant. Longitudinal trends in size of the Talek West clan and comparisons with other clans in the Reserve We analyzed mean monthly clan sizes to determine how the clan size of spotted hyenas in the Talek region has changed over time. Our goals here were twofold: 1) to investigate whether clan size in this region has changed over time, and 2) to determin e whether the Talek West clan was significantly different in size from the other clans in the Reserve. To do this, we first conducted a longitudinal analysis of Talek West, the clan exposed to high levels of anthropogenic disturbance in the Talek region, b y investigating trends in clan size from 1988 -2013 with year coded as a time -series variable. We then compared mean monthly clan sizes in Talek West to all other clans for which we had data in concurrent years. Mean monthly clan size data were available starting in 2002 for Mara River, 2007 for Fig Tree, and 2008 for Happy Zebra, Serena South, and Serena North. In both analyses, mean monthly clan sizes were overdispersed, so we utilized a negative binomial distribution. Trends in lion interactions with hye nas and lion sightings We performed longitudinal analyses to investigate temporal trends in lion numbers in the Talek region , and cross -sectional analyses to investigate differences among clans within the Reserve . We first modeled temporal trends in the mo nthly 54 frequency of the number of sightings when both lions and hyenas were present together to determine whether this changed significantly between 1988 and 2013. We modeled the number of hyena locations per month where lions were also present as a functio n of year coded as a time -series variable. We mean standardized the total number of locations at which hyenas were observed each month, and included this as an offset. Here we fit the model with a zero -inflation adjustment following Zipkin et al. (2010) , where the probability of inclusion of a month when no lions were sighted was mode led as a random variable. Second, we compared the mean monthly numbers of lions sighted in Talek during two five -year periods, 2004 -2008 and 2009-2013. We also compared the numbers of lions sighted in Talek in 200 9-2013 with the numbers of lions sighted in the territories defended by the Serena North, Serena South, and Happy Zebra clans of spotted hyenas during this same time period to identify differences that may be stemming from anthropogenic disturbance . To account for variation in the total amount of search effort invested in each territory, the number of hours each month during which observers were searching for carnivores within territory boundaries was mean-standardized and included as an offset. In both analyses, we utilized a negative binomial dist ribution because our counts were overdispersed. Spatial analysis of hyena and lion locations We modeled the distances at which lions and hyenas were found from the Reserve boundary in the Talek region as a function of hour and year coded as time series va riables, prey abundance, rainfall, and the number of livestock grazing in the Talek region during that month. We also included quadratic terms for each of these 55 covariates to investigate the possibility of nonlinear relationships. Year and hour were includ ed to investigate long -term trends and short -term daily patterns in these distances, respectively. For both models 1700 h was set as hour = 1. Locations at dens were not included in this analysis due to concerns regarding autocorrelation. These models were fit with a log-normal distribution. Results Temporal trends in the clan sizes of spotted hyenas Monthly clan sizes in the Reserve ranged from 22 to 126 individuals, with a mean ± SE of 55.26 ± 0.87 ( Figure 3 .2). The Talek West clan was consistently the l argest study clan; it also increased significantly in size over time ( Table 3.1) and underwent two clan fission events , in each of which the clan split from one parent clan into two discrete daughter clans ( Figure 3 .2; also see Holekamp et al. 1993) . These two fission events were completed by 1990 and 2001, respective ly, and both resulted in declines in overall clan size. Starting in 2008 the Talek West clan underwent rapid growth while sizes of the other clans studied in the Reserve remained relatively stable. The Talek West clan was significantly larger than the othe r study clans during nearly all years when concurrent data were available (2002 -2013; Table 3.1). 56 Figure 3.2. Trends in the sizes of spotted hyena clans in the Maasai Mara National Reserve, Kenya. The yearly mean ± standard error ( SE) clan sizes are pres ented for the Talek West (1988 -2013), Fig Tree (2007 -2013), Mara River (2002 -2013), Serena North (2008 -2013), Serena South (2008 -2013), and Happy Zebra clans (2008 -2013). Years without SE bars did not have multiple censuses. The two documented clan fission events in Talek West that occurred in 1989 and 2000 are indicated with horizontal lines. 20406080100120Year Clan total198819911994199720002003200620092012TalekSerena North Fig Tree Happy Zebra Serena South Mara River 57 Table 3.1. Summary of the mean posterior parameter estimates and 95 % Credible Intervals (CIs) for longitudinal trends in size of the Talek West clan of spotted hyenas (1988 -2013), and cross comparisons between the Talek West clan and other clans in the Reserve between 2002 and 2013. The Talek West clan has increased significantly over time, and was significantly larger than all other study clans. Significant post erior parameter estimates are indicated in bold font (95% CIs that did not overlap with zero). The parameter Year indicates the amount of change in each subsequent year. Parameter Mean posterior estimate (95% CI) Longitudinal trends in clan size of Talek West (1988 -2013) Mean monthly clan size 74.8 (72. 3, 77.37) Year 1.7 (1. 4, 2.02) Cross -comparisons of clan sizes of spotted hyenas Talek West (2002 -2013) 72.74 (69.5 9, 75.97) Mara River (2002-2013) 50.73 (41.6 6, 59.95) Fig Tree (2007 -2013) 49.02 (39.44, 59.61) Serena North (2008 -2013) 48.71 (41.0 6, 56.57) Serena South (2008 -2013) 40.6 (33.37, 47.9 2) Happy Zebra (2008 -2013) 31.77 (24.8 8, 39.06) Trends in lion interactions with hyenas and lion sightings Between 1988 and 2013, we recorded a total of 87,734 locations of spotted hyenas in the Talek region, and lions were also present at 1 ,188 of them. The number of interactions between lions and hyenas per month ranged from 0 to 18, with a mean ± SE of 3.92 ± 0.23. The frequency of interactions betw een lions and hyenas in Talek has significantly declined since 1988 (intercept and [95% CI]: 1.2 [0.1, 1.41]; change per year mean posterior estimate and [CI]: -0.03 [-0.04, -0.007]). We performed 315 monthly counts of lions within the territories of 4 stu dy clans between 2004 and 2013. 58 The total number of lions sighted ranged from 0 to 233 per clan territory per month, with a mean ± SE of 29.21 ± 2.08. There were significant differences in the mean numbers of lions sighted per clan ( Figure 3 .3). The number of lions sighted in the territory of the Talek West clan decreased significantly between 2004 -2008 (expected mean posterior [95% CI] = 34.9 2 [26.8, 46.39]) and 2009 -2013 (15.8 5 [12.002, 20.74]), to a level lower than that in Serena North (46.36 [35. 3, 60.58]), but not different from those in Serena South or Happy Zebra (13.0 2 [9.63, 17.38] and 18.75 [14.13, 23.98], respectively). 59 ! Figure 3.3. The expected mean number of lions sighted per month per clan of hyenas ± 95 % credible interval ( CI) between 2004-2008 and 2009-2013. The territory of the Talek West clan (TW) is located in the Talek region, and the territories of the Serena North (SN), Serena South (SS), and Happy Zebra (HZ) clans are located in the Mara Triangle. The expected mean count in TW sign ificantly declined between 2004 -2008 and 2009-2013 (posterior parameter estimates 95 % CIs did not overlap with one another). The expected mean monthly count of lions in TW in 2009 -2013 was significantly lower than SN, but no different than SS or HZ, durin g the same time period. ClanMean ± CI number of lions counted 0102030405060TWTWSNSSHZ2004-2008 2009-2013 60 Spatial analysis of hyena and lion locations Between 2004 and 2013, we recorded the locations of individual and groups of hyenas in the Talek West territory when the hyenas were not at dens a total of 23 ,470 times (mean ± SE locat ions per year: 2 ,347 ± 179.27), and recorded 1 ,161 locations of lions (mean ± SE locations per year : 116.1 ± 17.79). The distances at which hyenas and lions were found from the Reserve boundary changed significantly over time and were influenced by current ecological conditions ( Table 3.2). We observed significant increases in distances from the Reserve boundary over the course of this study among both lions and hyenas. The quadratic effect of year was also significant for lions and hyenas ( Table 3.2), indi cating that both species may have been decreasing their distances to the Reserve boundary in recent years. Lions and hyenas were more likely to be farther away from the edge of the Reserve during daylight observation hours than during the dark hours of observation before sunrise or after sunset ( Figure 3 .4), and also when prey were more abundant (Table 3.2) . The distances at which hyenas were found from the Reserve boundary were significantly predicted by the linear effects of rainfall and livestock; hyenas were more likely to be closer to the boundary during times of higher than average rainfall, and f arther away when livestock were grazing at higher numbers within the Reserve ( Table 3.2). Lions were more likely to be farther from the boundary of the Reserv e when rainfall was higher than average, but this was also a nonlinear relationship. The effect of livestock was not significant in predicting the distances at which lions were found from the Reserve boundary ( Table 3.2). 61 Table 3.2. Mean posterior summari es and 95 % Credible Intervals (CIs) for variables explaining the distances in meters at which spotted hyenas and lions were found from the Reserve boundary over time in the Talek region of the Maasai Mara National Reserve, Kenya. Locations of hyenas and l ions considered here were recorded from 2004-2013. Variables are listed in descending order of importance following year and hour based on the weights of their untransformed posterior parameter estimates. Statistically significant posterior parameter estim ates are indicated in bold font (95% CIs that did not overlap with zero) . Parameter Mean posterior estimate (95% CI) Hyenas Intercept 669.16 (644.78, 693.44) Year 61.88 (56.9 1, 66.75) Year2 -15.91 (-17.21, -14.56) Hour 98.91 (94.85, 103.01) Hour2 12.79 (12.35 , 13.23) Prey -48.62 (-58.25, -39.15) Livestock 25.44 (8.3 7, 42.96) Prey 2 13.91 (7.9 1, 19.85) Rain -10.78 (-19.18, -1.94) Rain2 -2.36 (-7.65, 2.73) Livestock 2 -1.61 (-6.26, 3.24) Lions Intercept 709.0 (592.08, 835.27) Year 37.87 (22.6 6, 55.0) Year2 -5.11 (-10.79, -0.17) Hour 32.22 (26.76, 37.84) Hour2 11.98 (10.71, 13.2 9) Prey 2 69.1 (40.78, 101.6 3) Prey -57.10 (-100.69, -14.23) Rain 51.08 (7.56, 98.37) Rain2 -33.53 (-64.48, -5.53) Livestock 54.19 (-20.64, 147.53) Livestock 2 -5.06 (-28.11, 16.73) 62 ! Figure 3.4. Mean ± standard error ( SE) distances at which spotted hyenas and lions were found from the Reserve boundary (a) over years of the study, and (b) over hours of the day during which we monitored lions and spotted hyenas . Discussion Our research presents new and important data documenting concurrent long -term trends in the populations of two long -lived species of large carnivores in the Maasai Mara National Reserve. The Reserve has long been seen as a stronghold for the con servation of large carnivores in east Africa (Ogutu & Dublin 2002; Riggio et al. 2013), yet our work suggests that the large carnivores inside this PA may not be adequately buffered from the anthropogenic acti vity occurring immediately outside the boundaries of the Reserve. In the Talek region at the edge of the Reserve, where thousands of livestock graze daily inside the park, and rapid habitat change is occurring Year Distance to edge (km) 0123420042006200820102012lionsspotted hyenas a.Hour01234b.567891718192021lionsspotted hyenas 63 immediately outside of Reserve boundaries, the local population of spotted hyenas has undergone rapid growth during the same period in which we observed a significant decline in the frequency of lion sightings. Two systematic assessments based on individual lion identities documenting the total number of lions inside the Reserve were conducted by Ogutu et al. (2002) and by Dloniak (2006) . DloniakÕs (2006) work indicated that the lion population had declined by 40% since Ogutu et al. (2002) estimated the lion population size 13 years earlier. Our research indicates th ese downward population trends in lions are likely to be continuing in the Reserve, at least in the Talek area. Although we do not have information on the total number or density of lions in the Talek region, both the total number of lions sighted in Tale k, and the frequency of their interactions with hyenas, declined significantly over time. These declines were concomitant with an increase in the size of the local clan of spotted hyenas. Thus, the effects of anthropogenic disturbance along the edge of the Reserve appear to be affecting lions and hyenas in radically different ways. However, in areas of the Reserve experiencing significantly less anthropogenic disturbance than the Talek region, such as the Mara Triangle, the patterns in clan sizes of spotted hyenas and lion sightings varied among clan territories. The Serena North territory had the largest clan size in the Mara Triangle, but also had the most lions sighted of all the clans studied there. Furthermore, it was in the territories of the smallest clans we monitored in the Mara Triangle where we observed numbers of lions sighted similar to those in Talek. Therefore, local ecological conditions also play a significant role in determining the numbers of these large carnivores (e.g. , prey availability) . 64 The decline in the number of lion sightings over time in the Talek region is unlikely to be due to decr eased detectability of lions in our study area . Lions and hyenas regularly occupy the same regions of the Reserve (Ogutu et al. 2005; Watts & Holekamp 2008; 2009) , and also show a high degree of spatial overlap throughout sub -Saharan Africa in environments where they both still occur (Creel & Creel 1996; P”riquet et al. 2014) . Based on searching for lions and hyenas during both hours of daylight and darkness during a ll years of this study, we infer that the apparent trends in lion sightings are real. Given the decline in lion sightings over time in the Talek region and the concurrent increase in the clan size of sympatric spotted hyenas, it appears that the anthropoge nic activities around Talek may be having a negative effect on lions and a positive effect on hyenas. In the current study we cannot infer causation between the increase in the size of the spotted hyena clan in Talek and the reduc ed frequency at which we s ighted lions, but it is possible that this may represent a case of mesopredator release. On the other hand, there are certain characteristics here that distinguish this from other documented cases of mesopredator release. For example, in contrast to most p airs of apex and meso -predators, lions and hyenas both prey primarily on medium- and large -bodied herbivores, and historically these species have exhibited a high degree of geographic overlap (Hayward & Kerley 2005; Hayward 2006; P”riquet et al. 2014). It could thus be argued that they offer similar ecosystem services through to p-down controls , which would distinguish them from the species pairs in other cases of mesopredator release . The long-term effects are currently unknown of a carnivore community in the Reserve that is made up of more spotted hyenas and fewer lions. To 65 our knowledge, no previous work describes a mesopredator release within a PA containing multiple large carnivore species living at high densities like we have here . Interestingly, the amount of rainfall and numbers of livestock grazing in Talek appear to be having different effects on the space -use of hyenas and lions. Whereas hyenas increased their distances from the boundary when higher than average numbers of livestock were grazing in this region, lions were not affected. Whereas higher than average rainfall increased the distance to the Reserve boundary for lions, it decreased this distance for hyenas. Heavy rainfall can cause prey to disperse, and bring carnivores into closer proximity to people, and thus ultimately increase rates of conflict (Kolowski & Holekamp 2006; Kissui 2008; Sogbohossou et al. 2011; Lyamuya et al. 2014) . In the current research, responses to rainfall appear to be bring ing hyenas closer to regions of conflict, while lions are moving farther away. The effects of prey were the same for lions and hyenas; an increase in prey availability decreased the distances at which both species were found from the Reserve boundary. Of the hours we sampled, both lions and hyenas were more likely to be closer to the boundary of th e Reserve during hours of darkness than daylight in Talek ( Figure 3.4). This is a novel finding for spotted hyenas, and supplements earlier evidence that lions prefer to be closer to areas containing dense human populations when human activity is expected to be the lowest (Schuette et al. 2013b; Oriol -Cotterill et al. 2015) . In contrast to previous research, our results show that this change is occurring inside Reserve boundaries. Our observation t hat both lions and hyenas were approaching Reserve boundaries more closely during hours of darkness than hours of daylight suggests that lions and hyenas might often be leaving the Reserve at night and moving 66 into the surrounding community lands where they may be preying upon livestock. GPS collar data from spotted hyenas and lions in Talek will be needed to test this hypothesis. However, conflict stemming from large carnivores leaving PAs to prey on livestock in community lands is a serious source of decli ning large carnivore populations worldwide , as it often results in retaliatory spearing, poisonings, and poaching (Woodroffe 2001; Woodroffe & Fra nk 2005; Ripple et al. 2014; Ogada 2014) . In the current study , these types of conflict may be responsible for the apparent decline in the Talek lion population. Above all, our research indicates that the carnivore populations within the Maasai Mara Natio nal Reserve are changing. If we wish to maintain this iconic Reserve as a stronghold for large carnivore populations, we need to perform a rapid assessment of the actual population size of lions, investigate in more detail the changes that are occurring wi thin the carnivore community and their underlying causal factors, and assess the effects of this altered carnivore guild on the ecosystem. PAs in sub -Saharan Africa are quickly becoming islands surrounded by degraded rangelands and exponentially growing hu man populations (Newmark 2008) . For large carnivores in particular, their future conservation relies on the effica cy of these sanctuaries to protect species, like lions, that may be particularly sensitive to habitat degradation and at high risk for conflicts with people (Packer et al. 2013; but see Creel et al. 2013a) . Successful commu nity conservation work can help to mitigate conflicts outside of PAs (Hazzah et al. 2014), and is integral to long -term conservation of large carnivores. However, if sensitive species continue to decline within PAs, even the best conservation efforts in community lands may not be able to sustain their populations into the future. 67 CHAPTER 4: Early warning signals of environmental degradation : Do the movements of spotted hyenas indicate the diversity and abundance of sympatric wildlife? 68 Introduction Worldwide human population growth has become a major threat to global biodiversity (McKee et al. 2003; Butchart et al. 2010) . The direct (e.g. , bushmeat consumption) and indirect (e.g. , climate change, competition for resources) effects of human population growth on native wildlife occur in nearly all ecosystems, an d have led to many long -term declines in wildlife populations. One mechanism to help understand ecological changes associated with human population growth is the use of indicator species , which can be used as prox ies to reveal ecosystem trends, offer insig ht into appropriate conservation strategies, and vastly improve conservation and management decisions (Landres et al. 1988; Caro & O'Doherty 1999; Caro 2010) . Indicator species have been used in the past as proxies to assess changing environmental conditions in many ecosystems (Landres et al. 1988; Caro & O'Doherty 1999; Caro 2010) . Whereas use of indicator species was hailed early o n as a n important surrogate measure to assist in wildlife conservation, recent research has cast doubt on their utility (Fleishman & Murphy 2009; Kessler et al. 2011) . When selecting an indicator species to understand broad ecological trends, the chosen species should occur throughout a wide geographic range, be available for continuous assessment, be cost -effective and relatively easy to measure, have a well -understood natural history, and respond quickly to ecological change (Noss 1990; McGeoch & Chown 1998) . Other recommendations are that indicator species should be large -bodied, have large home ranges, and occupy high trophic positions so they can reflect ecological change at multiple trophic levels (Caro & O'Doherty 1999) . Nevertheless, many organisms previously classified as indicator species lack some or 69 all of these qualities (Hilty & Merenlender 200 0). Furthermore, indicator species have traditionally been used on a presence/absence basis to assess the effects of environmental change on other sympatric species. That is, if the indicator species is present in an ecosystem, then you are likely to find other sympatric species there as well (Fleishman et al. 2005; Caro 2010) . Thus traditional indicator species have been selected for their sensitivity to specific types of environmental change. For example, common farmland birds in Europe have been proposed as indicator species for the effects of agricultur e on the biodiversity of sympatric wildlife because of their sensitivity to agricultural production (Gregory et al. 2005) . If these birds are absent or declining , this indicates other wild birds in Europe are also likely to be declining. H owever, long -lasting d amage may occur before this type of an indicator species can be identified and monitored. Additionally, by the time these sensitive species are eradicated, long -lasting damage may already have been done to sympatric wildlife populations. Therefore, it would be useful to find an indicator species that can indicate impending changes to populations of sympatric species through more labile metrics (Caro 2010) , such as behavioral measures . Behavioral changes in one species coul d then be used to better understand trends in sympatric wildlife. Here we inquired whether the movement patterns of spotted hyena ( Crocuta crocuta ) could indicate changes in populations of sympatric mammals. Spotted hyenas exhibit many of the traits sought in idealized indicator species as described above (e.g., wide distribution, well-understood natural history, high trophic position, respond quickly to ecological change , etc. ). Having an indicator species that can indicate the abundance and diversity of sympatric species 70 before any changes become apparent would potentially be extremely useful for wildlife managers and conservationists throughout the African continent. The spotted hyena has a remarkably high degree of intraspecific variation in behavior an d occurs over a broad geographic range (Holekamp & Dloniak 2010) . Spotted hyenas are extremely adaptable carnivores that can survive by scavenging on carrion , but they are also very efficient hunters of live prey. In contrast to most other African animals occupying similar trophic positions, spotted hyenas thrive on a wide diversity of prey species (Kruuk 1972; Cooper et al. 1999; Hayward 2006) , breed year round (Holekamp et al. 1996), and can be active either day or night (Kruuk 1972; Kolowski et al. 2007; Stratford & Stratford 2011) . Also, spotted hyenas can persist at high densities even in areas of intensive human activity (e.g. , Ethiopia) where all of their natural prey have been extirpat ed (Yirga et al. 2012; 2013) . The space -use behaviors of spotted hyenas exhibit a number of important characteristics that may allow them to inform us about deteriorating ecosystems while there is still time to repair damage inflicted by anthropogenic activity. Previous research suggested that the behavior of hyenas changes in response to anthropogenic activity roughly 5 years before any changes become apparent in their population dynamics (Kolowski & Holekamp 2009) . Thus, these animals emit early warning signals about their own future welfare. Here our goal was to determine whether the behavior of spotted hyenas might also correlate with certain characteristics of sympatric wildlife. We tested the hypothesis that the spotted hyena can serve as an indicator species in the northernmost portion of the Mara -Serengeti ecosystem, within the Maasai Mara National Reserve (henceforth, the Reserve), in southwestern Kenya. Specifically, 71 we were interested in examining how the behaviors of spotted hyenas may correlate with numbers and diversity of sympatric wildlife. The Reserve has historically supported a great abundance and diversity of carnivores, resident and migrant herbivores, and many species vulnerable to extinction including black rhinos ( Diceros bicornis ) and African elephants ( Loxodonta africana ). However, some areas of the Reserve are currently und ergoing massive ecological changes associated with increasing anthropogenic activity along Reserve borders. Human population growth around the Reserve has increased exponentially since 1950 (Lamprey & Reid 2004 ; see also Chapter 2 of this dissertation ), resulting in habitat degradation due t o mechanized agriculture (Serneels et al. 2001; Lamprey & Reid 2004; Ogutu et al. 2009) , urbanization, and the intensive grazing of lives tock by pastoralist communities (Thompson & Homewood 2002; Ogutu et al. 2005). Evidence suggests that these threats may be responsible for long -term declines in populations of resident and migrant herbivores (Ottichilo et al. 2000; Homewood et al. 2001; Serneels & Lambin 2001; Ottichilo et al. 2001; Ogutu et al. 2005; Norton -Griffiths et al. 2008; Ogutu et al. 2009; 2011 ; Chapter 2 of this dissertation ), behavioral and physiological changes in spotted hyenas (Boydston et al. 2003a; Kolowski et al. 2007; Kolowski & Holekamp 2009; Van Meter et al. 2009) , and altered population sizes of resident carnivores within Reserve boundaries (Ogutu et al. 2005; Dloniak 2006 ; Chapter 3 of this dissertation ). By using naturally -occurring variation in ecological change stemming from proximity to anthropogenic disturbance, we inquired whether the movements of spotted hyenas were correlated wit h the abundance and diversity of sympatric wildlife populations, and if so, how far in advance. Examining the effects of anthropogenic disturbance and ecological change in a complex 72 system like the Mara -Serengeti is quite challenging ; an indicator species might assist conservation and management officials in understanding how this ecosystem may be faring over time. To test our hypothesis, w e monitored the fine -scale movement parameters of speed and the tortuosity of paths taken by spotted hyenas fitted with GPS collars, and combined these measurements with in -depth sampling of sympatric herbivore and carnivore populations within the Reserve. The speed of movement and turning angle of large carnivores are known to be influenced by anthropogenic and ecological factors (e.g., lions (Panthera leo ); Oriol -Cotterill et al. 2015) . Thus, we predict ed that these movement parameters of spotted hyenas w ould also be influenced by current ecology, and most importantly, and that they might also be able to serve as an indicator of the abundance and diversity of sympatric wildlife. By using the naturally -occurring variation in anthropogenic disturbance and ecological conditions within the Reserve, we first investigated the ecological and anthropogenic factors that influence the speed and tortuosity of travel by collared spotted hyenas. We then asked to what extent these movement patterns of spotted hyenas correlate with the abundance and diversity of sympatric species at future time points . Methods Study site The Reserve is comprised primarily of open grassland interspersed with riparian habitat along seasonal watercourses. The Reserve has traditionally supported many resident herbivore and carnivore species year -round, and is also visited seasonally by 73 migrating wildebeest ( Connochaetes taurinus ), plains zebra ( Equus quagga ), and ThomsonÕs gazelle ( Eudorcas thomsonii) populations from the Serengeti National Park to the south (Bell 1971; Sinclair & Norton -Griffiths 1979) , and the Loita Plains to the northeast (Stelfox et al. 1986) . The only permanent river in this ecosystem, the Mara River, serves as the boundary between two different management agencies (Figure 4.1). 74 ! Figure 4.1. Locations of the study clans and loops (dotted lines) run to sample sympatric wildlife in the Maasai Mara National Reserve, Kenya. The clans in areas managed by the Mara Conservancy are Serena North (SN) and Serena South (SS). The clan of hyenas in the region exposed to severe anthropogenic disturbance, and managed by the Narok County Government, is the T alek West clan (TW). Talek West is located in the Talek region of the Reserve. Territory boundaries are indicated with thick dashed lines. The locations of the weather stations ( ) and the town (!) closest to the TW territory (Talek) are also noted. !"#"$%&'(# 75 The Narok County Government (NCG) manages the eastern side of the Reserve . Within this portion of the Reserve (Figure 4.1), the Talek region , in particular , is an area near the Reserv e border that in recent decades has been subjected to a great deal of anthropogenic disturbance. This disturbance include s increases in the number of human settlements and tourist facilities outside the Reserve, as well as an exponential increase in numbers of livestock grazing inside Reserve boundaries (Boydston et al. 2003a; Lamprey & Reid 2004; Kolowski & Holekamp 2009 ; Chapter 2 of this dissertation ); here we refer to these effects collectively as Òanth ropogenic disturbance .Ó In contrast to the Talek region, other portions of the Reserve are far less affected by anthropogenic disturbance. In particular, the portion of the Reserve west of the Mara River, called Òthe Mara Triangle,Ó is remarkably pristine. The Mara Conservancy (TMC) manages The Mara Triangle, and the wildlife in this area encounter s virtually no anthropogenic activity at all aside from visitation by tour vehicles. We were able to exploit this naturally -occurring variation in anthropogenic d isturbance to investigate the ecological and anthropogenic influences on behavior by spotted hyenas, and determine whether and how the behavior of spotted hyenas might correlate with numbers of sympatric wildlife in both disturbed and undisturbed areas. Monitoring the movements by spotted hyenas Spotted hyenas are gregarious, large carnivores that live in social groups called clans. Each clan is comprised of immature juveniles, adult females, and immigrant males. Females are philopatric except in the rare cases of clan fission events (Holekamp et al. 1993) . Males emigrate from the ir natal clan upon reachi ng sexual 76 maturity (van Horn et al. 2003) . On average, clans of spotted hyenas are comprised of 29 individuals throughout sub -Saharan Africa (Holekamp & Dloniak 2010) , and clan members defend a communal territory by scent marking and sparring with neighborin g clans (Kruuk 1972; Mills 1990b; Henschel & Skinner 1991; Boydston et al. 2001) . Each clan is structured by a linear matrilineal dominance hierarchy in which natal animals dominate immigrant males; immigrant males hold ranks relative to one another based on their tenure in the clan. The space -use of spotted hyenas can be influenced by many factors, including rainfall, prey availability, social rank, time of day, and indirect effects stem ming from anthropogenic disturbance by cattle grazing (Kruuk 1972; Mills 1990b; Boydston et al. 2003a; 2003b; Kolowski et al. 2007; Kolowski & Holekamp 2009; Stratford & Stratford 2011). In areas where prey are relat ively abundant, spotted hyenas travel an average of 11 km per day (Kruuk 1972; Kolowski et al. 2007) , but have been documented to travel up to 75 km per day in areas with fewer prey (Henschel & Skinner 1987; Mills 1990a; Hofer & East 1993) . Spotted hyenas in the Reserve are capable of crossing their entire territory (~10km) in < 1 hour (Boydston et al . 2003a). We studied three clans of spotted hyenas to investigate ecological and anthropogenic influences on their movements (Table 4.1), and to see if the movements exhibited by individual group members were correlated with the future abundance and diversity of sympatric wildlife. Two clans were located in the relatively undisturbed Mara Triangle, managed by TMC ( ÒSerena North Ó and ÒSerena South Ó), and one was located in the highly disturbed Talek region on the eastern side of the Reserve, managed by NCG (ÒTalek West Ó; Figure 4.1). Animals were individually identified by their spot 77 patterns, ear damage and other unique features. Standardized ranks were calculated for all natal clan members ranging from 1 to -1 based on outcome s of dyadic aggressive interac tions (Holekamp & Smale 1993) ; a value of 1 was assigned to the highest -ranking female in the clan and a value of -1 to the lowest. Table 4.1. Salient characteristics of the Talek West, Serena North, and Serena South clans of spotted hyenas monitored in the current study in 2013. Standard errors are presented in parentheses. The number of tourists per territory was estimated as the number of vehicles counted biweekly during count s by Mara Hyena Project personnel . Parameter Talek West Serena North Serena South Clan size (no. hyenas) 113.00 (1.80) 51.42 (0.87) 39.25 (1.34) Territory size (km 2) 77.04 42.67 28.29 Mean density of hyenas 1.47 / km2 1.21 / km 2 1.39 / km 2 Tourists (no. encountered) 8.18 (1.79) 13.28 (1.76) 1.89 (0.64) Livestock (no. inside territory) 2218.85 (240.35) 0 0 We anesthetized 6 -8 spotted hyenas in each clan with Telazol (6.5 mg/kg) administered in a plastic dart fire d from a CO 2 powered rifle (Telinject Inc., Saugus, California), and fitted them with GPS radio collars (Vectronic Aerospace, Berlin, Germany) to monitor fine -scale movement behaviors. Collars were deployed on parous females of low - and high-ranks (i.e. , only in the lowest and highest third s of the rank hierarchy for that clan, respectively). We set the collars to record GPS locations of the hyenas and the ambient temperature at hourly intervals from 1600 h to 1000 h, and once again at 1300 h. Speeds and tu rning angles (indicating the degree of tortuosity) 78 were calculated between successive hourly locations in the movement paths of the hyenas with the package ÔadehabitatLTÕ in R (Calenge 2006; R Core Team 2015) . For all analyses, we used the absolute value of the turning angle as a measure of tortuosity, regardless of the directionality of each hyenasÕ trajectory. Sampling the abundance of sympatric wildlife A grid of 200 x 20 0 m cells was delineated on a map of each hyena territory, and we calculated the estimated abundance of each of the following animals within each 200 x 200 m grid cell (henceforth, cell) every 4 -6 weeks between July 2012 -March 2014: olive baboon (Papio anu bis), banded mongoose (Mungos mungo ), bat-eared fox (Otocyon megalotis ), black -backed jackal (Canis mesomelas ), buffalo (Syncerus caffer ), bushbuck ( Tragelaphus scriptus ), caracal (Caracal caracal ), cheetah ( Acinonyx jubatus ), KirkÕs dik dik ( Madoqua kirki i), eland (Taurotragus oryx ), African elephant, Maasai giraffe ( Giraffa camelopardalis tippelskirchi ), GrantÕs gazelle (Nanger granti ), hartebeest (Alcelaphus buselaphus ), hippo (Hippopotamus amphibius ), spotted hyena, impala (Aepyceros melampus ), leopard (Panthera pardus ), lion ( Panthera leo ), oribi ( Ourebia ourebi ), ostrich (Struthio camelus ), bohor reedbuck ( Redunca redunca) , black rhino, serval ( Leptailurus serval ), side -striped jackal ( Canis adustus ), slender mongoose (Galerella sanguinea ), ThomsonÕs g azelle, topi ( Damaliscus korrigum jimela) , vervet monkey (Chlorocebus pygerythrus ), warthog (Phacochoerus africanus ), wildebeest, and zebra. To estimate the number of animals in each cell, we employed a novel sampling technique that exploits the easy obs ervability of animals in the ReserveÕs open 79 grassland habitat. We first defined sampling loops (henceforth, ÒloopsÓ) within each territory to allow for maximum coverage and minimal opportunities for double counting wildlife (Figure 4.1). All efforts were m ade to include > 500 m of track in each square kilometer of each territory, and resulted in 35.28 Ð44.52 km of loop in each of the three territories we monitored. Part of the territory of the Talek West clan extends outside Reserve boundaries; we also inclu ded a separate 8.95 km loop in this region (Figure 4.1). All loops were sampled on preexisting roads and tracks. Loops were sampled on each of three consecutive days ( each of these three days was called a Òreplicate Ó) in each territory every four to six we eks ( each three -day sampling period was called a Òsample Ó), unless limited by weather. We randomly chose three starting locations on each loop equidistant from one another. Each replicate in a sample started at a different location in an attempt to capture any temporal variation in herbivore movement throughout the day. Observers (1-3) started each loop per sample at first light by driving our vehicle along the loopÕs centerline. Upon first detecting an animal or group of animals, we recorded the GPS locati on of our vehicle, the distance to the animal(s) from the vehicle using a laser rangefinder (Nikon Laser 1200), and the bearing to the animal(s) from the vehicle. Using these data, we projected the GPS location of the animal(s). Animals of the same species that were separated by < 200 m were considered to be part of the same grouping, and were assigned to one location at the center of the group. To account for effects of sampling error due to variation in the observability of wildlife, we calculated species -specific detection probabilities using the package ÔunmarkedÕ in R v. 3.2.0 (Fiske & Chandler 2011) . Detection probabilities were assumed 80 to decay as a function of the distance from the centerline of the loop (Buckland et al. 2001). For all species with > 100 locations, we modeled the probability of detection as a function of vehicle type (small or large) in which the replicate was performed to account for variation in detection due to grass height , and the number of observers in the vehicle (1 or > 1). We used hazard functions because they had the lowest AIC values in fitting our data (Buckland et al. 2001). Using the models with the best fit, we determined the distance for each species at which its detection probability equaled or exceeded 90 %. The average ± standard error 90 % detection distances for herbivores and carnivores with more than 1 00 sightings were 393.61 ± 38.27 and 140.02 ± 14.51 m, respectively. The 90 % detection probability of species with fewer than 100 locations was assumed to be 100 m from the centerline of the loop. We then buffered the loops by each of these species -specif ic detection distances to create a detection region for each species in each replicate . Locations of animals we sighted during replicates that fell within these detection regions were then assigned to the cell in which they were located, within that replic ate of that sample, and were used in all analyses. Observer effort per cell per species was defined as the number of square meters in which the detection region for that species overlapped with the cell. Statistical analyses Predictors of movement pattern s by spotted hyenas To investigate the ecological and anthropogenic influences on the movement patterns of spotted hyenas, we used linear mixed -effects models fit by maximum likelihood in the ÔnlmeÕ package in R v. 3.2.0 (R Core Team 2015) . We modeled the 81 turning angle and speed between successive locations of spotted hyenas as a function of clan membership, rank (high or low), whether or not the location was inside or outside of the Reserve, and both h our and hour 2 fit as time -series variables. We hypothesized that the speed between successive locations of spotted hyenas would be a function of their turning angles, and similarly, that the turning angles between successive locations would be a function o f their speeds ; we included each of these predictor variables in our models to investigate the interactions between these movement parameters. We also hypothesized that intra -clan competition and time of day would influence the movement behaviors of spotte d hyenas, and that these effects would vary among clans. Low -ranking hyenas could alter their movement behaviors to avoid conflict with high -ranking conspecifics, for example, by moving faster than high -ranking hyenas or by moving during different hours of the day. Movement behaviors of spotted hyenas were also hypothesized to vary with exposure to anthropogenic disturbance; we predicted that spotted hyenas in the Talek West clan would move faster than Serena North or Serena South hyenas to avoid conflicts with livestock herders, and that low -ranking hyenas would be more strongly affected than high -ranking hyenas. Therefore, we fit interactions between clan and rank, clan and hour, and hour and rank in our models. Rainfall, ambient temperature, and moonlight were also hypothesized to influence the movement patterns of spotted hyenas and were therefore included in our models. The GPS collars recorded ambient temperature hourly . The amount of daily rainfall in mm was collected at two weather stations in the Res erve (Figure 4.1). Precipitation was recorded daily at 0500 h; we modeled the speed and turning angles of hyenas during any given night (1600 h to 1000 h) as a function of data from the 82 following morning at 0500 h. We assigned a value representing the prop ortion of the moon that was illuminated for movement behaviors that occurred after sunset and moonrise, but before moonset or sunrise. All moons visible between these hours were assigned a value between 0 and 1 corresponding to the proportion of the moon t hat was illuminated on that night (i.e. , full moon = 1, new moon = 0). We assigned a moon value of 0 to hours in which the moon was not present, or if the movement value was calculated before moonrise on that day. The percentage of moon illuminated each ni ght was obtained from the U.S. Naval Observatory (http://aa.usno.navy.mil/data/docs/MoonFraction.php). To account for any variation in the speed of movement or path tortuosity among the spotted hyenas monitored in this study not explained by our other cova riates, we nested each nightÕs identity in hyena identity, and coded it as a random effect. This nesting also account ed for any additional interactions between the day on which the movement parameters were collected and hyena identity. Temporal serial auto correlation was accounted for by fitting both models with first -order autoregressive covariance structures. Speed between successive points was log -transformed to achieve normality, but turning angles were normally distributed. We ran one full model with a ll parameters included, separately, for both speed and turning angle. Spotted hyenas as an indicator species We analyzed the movements of spotted hyenas to investigate whether and how they might correlate with the local abundances of sympatric wildlife wi thin Reserve boundaries in the future . To do this, we used the ÔglmmADMBÕ package in R v. 3.2.0 83 (Bolker et al. 2012; R Core Team 2015) . We calculated the total number of resident herbivores, migrant herbivores, carnivores, and the value of species richness in each 200 x 200 m cell per sample. The abundance and diversity of wildlife in each cell were then modeled as a function of the speeds and turning angles exhibited by spotted hyenas in that same cell before the sampl ing event. We investigated how the movement of hyenas over a course of two weeks might correlate with the abundance an d diversity of sympatric wildlife in our samples at five time points: 2, 8, 14, 20, and 26 weeks later. We did not investigate time periods beyond 26 weeks due to the high degree of seasonality in herbivore movements and weather patterns in this environme nt. All turning angles and speeds of movement were mean -standardized. We also included social rank to investigate how movements of low- and high-ranking hyenas might correlate differently with the abundance and diversity of sympatric wildlife. We expected the movements of spotted hyenas in pastoralist community lands to be influenced by additional factors which might alter the relationship between their movements and the populations of sympatric wildlife (e.g., threats of conflict, range contraction, habitat change; Ogutu et al. 2011; Elliot et al. 2014; Oriol -Cotterill et al. 2015 ). Therefore, w e only used locations of spotted hyenas within Reserve boundaries for the indicator species analyses to avoid these confounding factors . The total amount of observer effort in each square meter within each cell per sample was mean -standardi zed and included as an offset. All of our dependent variables were zero -inflated and overdispersed; we fit these models with a negative binomial distribution and a constant zero -inflation parameter. Models for each time point were run separately for each dependent variable . Thus, for each time -point we ran one full model with speed, angle, 84 and rank. The first model in the series of time -points in which a behavior emitted by monitore d hyenas failed to significantly correlate with the abundance or diversity o f sympatric wildlife was considered to be the cessation of any relationship between the movements of hyenas and that ecological factor. Subsequent models were then run with that behavior removed. We present results from the models in which a hyena movement behavior significantly correlated with species richness or mammalian herbivore or carnivore abundance ( P < 0.05). Results Predictors of the movement patterns by spotted hyenas From 22 adult female hyenas w e collected a total of 138,751 locations that wer e separated by one hour (mean number of locations ± SE per hyena: 6306.86 ± 619.40) on 953 nights (mean number of nights ± SE per hyena: 498.41 ± 38.45) between April 2012 and November 2014. There was no significant difference between Serena North and Sere na South clans in our models ( P > 0.10), so we combined data from these two clans in all subsequent analyses and created the variable ÒManagement agency Ó to contrast NCG with TMC. Turning angles of spotted hyenas were predicted by the speed at which spotte d hyenas traveled, the time of day , whether or not the hyena was located inside or outside Reserve boundaries , and daily rainfall (Table 4.2) . None of our other model parameters significantly predicted the turning angles of spotted hyenas ( P > 0.05, Table 4.2). The speed at which spotted hyenas traveled between locations was significantly predicted by turning angle, time of day , ambient temperature, daily rainfall, social rank, 85 management agency, whether or not hyenas were located inside or outside the Rese rve , moonlight, an interaction between management agency and time of day , and an interaction between management agency and the social rank of the hyena (Figure 4.2 , Table 4.2). Post -hoc tests indicated the differences in management agency , and the interact ions between management agency and rank and management agency and time of day, were significant even when only looking at locations of spotted hyenas within Reserve boundaries (management agency within Reserve: estimate ± SE = -0.744 ± 0.114; P # 0.001; Mara Conservancy x low rank within Reserve: estimate ± SE = 0.4 4 ± 0.17; P = 0.0 19; Mara Conservancy x hour within Reserve : estimate ± SE = -0.024 ± 0.004; P # 0.001). The interaction between social rank and hour had no effect on the speed of move ment by hyenas ( P > 0.05 ; Table 4.2). 86 Figure 4.2. Speeds of movement by spotted hyenas inside Reserve boundaries of the Maasai Mara National Reserve, Kenya, in areas managed by the Narok County Government ( ) and the Mara Conservancy ( !). (a) Mean ± sta ndard error ( SE) speeds by individuals throughout the night. Hours of darkness are shaded in grey. (b) Mean ± SE speed of movement by spotted hyenas by rank and management agency . 0.00.10.20.30.4HourSpeed (m/s) a.17192123135790.160.170.180.190.200.210.22Speed (m/s) 0.160.170.180.190.200.210.22HighLowSocial rank b. 87 Table 4.2. Results from the generalized linear mixed -effects models that pr edicted the speed and turning angles of spotted hyenas between successive locations separated by one hour inside and around the Maasai Mara National Reserve, Kenya. Parameter estimates, Standard Errors (SEs), Degrees of Freedom (DF), T-, and P-values are displayed. Significant parameter estimates are indicated in bold font (P < 0.05). Parameter Value (SE) DF T P-value Speed Intercept -2.197 (0.1) 126529 -22.068 < 0.001 Turning angle -0.266 (0.006) 126529 -45.13 < 0.001 Management agency (TMC) -0.686 (0.114) 18 -6.026 < 0.001 Rank (Low) -0.321 (0.134) 18 -2.397 0.0276 Hour -0.257 (0.004) 126529 -61.705 < 0.001 Hour2 -0.031 (0.001) 126529 -59.378 < 0.001 Reserve (Inside) -0.314 (0.034) 126529 -9.16 < 0.001 Temperature -0.207 (0.01) 126529 -19.866 < 0.001 Rain 0.019 (0.009) 126529 2.062 0.0392 Moonlight 0.14 (0.024) 126529 5.765 < 0.001 Management agency x Hour -0.021 (0.004) 126529 -5.826 < 0.001 Management agency x Rank 0.431 (0.17) 18 2.533 0.0208 Rank x Hour 0.005 (0.004) 126529 1.301 0.1934 Turning angle Intercept 1.712 (0.025) 126529 69.424 < 0.001 Speed -0.669 (0.011) 126529 -59.484 < 0.001 Management agency (TMC) 0.005 (0.027) 18 0.171 0.8663 Rank (Low) -0.06 (0.032) 18 -1.904 0.073 Hour 0.005 (0.001) 126529 3.801 < 0.001 Hour2 0.001 (0.001) 126529 2.087 0.0369 Reserve (Inside) -0.079 (0.012) 126529 -6.752 < 0.001 Temperature -0.006 (0.004) 126529 -1.68 0.0929 Rain 0.006 (0.003) 126529 2.102 0.036 Moonlight 0.003 (0.008) 126529 0.42 0.674 Management agency x Hour 0.001 (0. 001) 126529 -0.388 0.698 Management agency x Rank 0.03 (0.04) 18 0.75 0.463 Rank x Hour -0.002 (0.001) 126529 -1.448 0.148 88 Spotted hyenas as an indicator species We conducted 16 samples of the abundance and diversity of herbivores and carnivores in t he Talek West, Serena North, and Serena South territories between July 2012 and March 2014. This sampling resulted in a total count of 364,905 individual animals with a mean ± SE of 4.956 ± 0.269 per cell. Estimated counts of wildlife were calculated in 5, 039 cells inside and outside of the Reserve (Figure 4.3). The average ± SE numbers of resident and migrant herbivores, and carnivores counted per grid cell were 2.334 ± 0.087, 2.933 ± 0.277, and 0.037 ± 0.003, respectively . This varied depending on whether sampling was done inside or outside the Reserve (resident herbivores inside: 2.507 ± 0.093; resident herbivores outside 0.001 ± 0.001; migrant herbivores inside: 3.133 ± 0.296; migrant herbivores outside: 0 .001 ± 0.001; carnivores inside: 0.039 ± 0.003; carnivores outside : 0.001 ± 0.001). Average ± SE number of species found per cell was 0. 411 ± 0.012, which also differed depending on whether sampling was done inside or outside of the Reserve ( species richness inside: 0.434 ± 0.012; species richness outsid e: 0.001 ± 0.001). Sample sizes for speeds of movement and tortuosity values of hyenas within cells for which sympatric wildlife information was available varied among 2 -, 8-, 14-, 20-, and 26-week time points (2 -week n = 39,966, 8 -week n = 35,122, 14 -week n = 33,010, 20 -week n = 29,830, 26 -week n = 26,961). 89 Figure 4.3. Variation in the abundance and distribution of (a) resident (b) and migrant herbivores, (c) carnivores, and (d) species richness within the Serena North (SN), Serena South (SS), and Talek W est (TW) territories in the Maasai Mara National Reserve, Kenya. Darker shades indicate a comparatively greater abundance (a -c), or diversity (d) in that 200m x 200m grid cell than in lighter cells during this study, accounting for effort. Open areas withi n territory boundaries indicate regions with no sampling effort. 90 91 Cells in which spotted hyenas moved faster were correlated with fewer carnivores and herbivores , and with lower species richness at future time points . This significant trend occurre d at the 2-, 8-, and 14- week time -points for resident herbivores, the 2-, 8-, 14-, 20-, and 26-week time -points for species richness, and at the 2 -, 8-, 14-, and 20-week time -points for carnivores and migrant herbivores ( P < 0.05; Table 4.3). Turning angl e was a significant negative predictor of resident herbivore abundance 2 weeks later, and species richness 2, 8, 14, 20, and 26 weeks later. Turning angle was a significant positive predictor of migrant herbivore abundance 2 and 8 weeks later, and of carni vore numbers 2 weeks later ( P < 0.05; Table 4.3). Rank of the hyenas was a significant predictor of species richness and the numbers of sympatric wildlife. Cells in which hyenas of low rank were more likely to be found had higher numbers of resident 92 and migrant herbivores, and greater species richness ( P < 0.05), but fewer carnivores (P < 0.05), than did cells in which hyenas of high rank were found ( Table 4.3). Table 4.3. Results from the generalized linear mixed -effects models that predicted herbivore an d carnivore abundance, and species richness, as a function of the speed, turning angles, and rank of spotted hyenas in the same area previous to the count. The behaviors of spotted hyenas significantly predicted the numbers of resident herbivores 2, 8, and 14 weeks before the sample, species richness 2, 8, 14, 20, and 26 weeks before the sample, and numbers of migrant herbivores and carnivores 2, 8, 14, and 20 weeks before the sample. Significant parameter estimates are indicated in bold font (P < 0.05). Parameter Estimate (SE) Z P Resident herbivores 2 weeks Intercept 1.13 (0.019) 59.680 # 0.001 Speed -0.034 (0.009) -3.600 ! 0.001 Angle -0.044 (0.01) -4.450 ! 0.001 Rank (low) 0.218 (0.019) 11.630 ! 0.001 8 weeks Intercept 1.164 (0.021) 56.160 # 0.001 Speed -0.055 (0.01) -5.340 ! 0.001 Angle -0.02 (0.011) -1.860 0.062 Rank (low) 0.167 (0.02) 8.170 ! 0.001 14 weeks Intercept 1.112 (0.021) 52.980 # 0.001 Speed -0.049 (0.01) -4.660 ! 0.001 Rank (low) 0.104 (0.021) 4.940 ! 0.001 Specie s richness 2 weeks Intercept -1.295 (0.011) -115.450 # 0.001 Speed -0.057 (0.008) -6.740 ! 0.001 Angle -0.032 (0.009) -3.680 ! 0.001 Rank (low) 0.138 (0.017) 8.250 ! 0.001 8 weeks Intercept -1.361 (0.013) -105.770 # 0.001 93 Table 4.3 contÕd Speed -0.054 (0.009) -5.730 ! 0.001 Angle -0.024 (0.01) -2.420 0.016 Rank (low) 0.174 (0.019) 9.280 ! 0.001 14 weeks Intercept -1.325 (0.013) -100.300 # 0.001 Speed -0.062 (0.01) -6.300 ! 0.001 Angle -0.026 (0.01) -2.590 0.010 Rank (low) 0.089 (0.019) 4.550 ! 0.001 20 weeks Intercept -1.427 (0.015) -95.250 # 0.001 Speed -0.012 (0.011) -1.090 0.274 Angle -0.043 (0.011) -3.840 ! 0.001 Rank (low) 0.117 (0.022) 5.450 ! 0.001 26 weeks Intercept -1.395 (0.016) -87.050 # 0.001 Angle -0.073 (0.012) -6.110 ! 0.001 Rank (low) 0.08 (0.023) 3.440 ! 0.001 Carnivores 2 weeks Intercept -2.388 (0.046) -52.330 # 0.001 Speed -0.386 (0.036) -10.790 ! 0.001 Angle 0.094 (0.029) 3.290 0.001 Rank (low) -0.75 (0.064) -11.680 ! 0.001 8 weeks Intercept -3.491 (0.066) -53.000 # 0.001 Speed -0.158 (0.043) -3.660 ! 0.001 Angle -0.012 (0.042) -0.300 0.767 Rank (low) -0.122 (0.082) -1.480 0.140 14 weeks Intercept -3.356 (0.063) -53.690 # 0.001 Speed -0.258 (0.047) -5.560 ! 0.001 Rank (lo w) -0.736 (0.093) -7.920 ! 0.001 20 weeks Intercept -3.477 (0.073) -47.460 # 0.001 Speed -0.138 (0.047) -2.960 0.003 Rank (low) -0.219 (0.091) -2.400 0.016 Migrant herbivores 2 weeks Intercept 0.493 (0.042) 11.610 # 0.001 Speed -0.114 (0.0 19) -5.960 ! 0.001 Angle 0.042 (0.018) 2.320 0.020 94 Table 4.3 contÕd !Rank (low) 0.531 (0.035) 15.340 ! 0.001 8 weeks Intercept 0.682 (0.047) 14.590 # 0.001 Speed -0.146 (0.022) -6.820 ! 0.001 Angle 0.085 (0.02) 4.330 ! 0.001 Rank (low) 0.412 (0.0 38) 10.970 ! 0.001 14 weeks Intercept 0.81 (0.046) 17.550 # 0.001 Speed -0.086 (0.021) -4.160 ! 0.001 Angle 0.028 (0.02) 1.400 0.160 Rank (low) 0.182 (0.038) 4.760 ! 0.001 20 weeks Intercept 0.696 (0.054) 12.860 # 0.001 Speed -0.07 (0.023) -3.020 0.003 Rank (low) 0.141 (0.044) 3.200 0.001 Discussion Here we present results showing that the movement patterns of spotted hyenas correlate with the future abundance and diversity of sympatric wildlife. Previous research suggested the space -use be haviors and stress physiology of spotted hyenas can be used to indicate the hyenasÕ own future welfare (Kolowski & Holekamp 2009; Van Meter e t al. 2009). The results from our study show that the movement patterns of spotted hyenas also correlate with the abundances and diversity of sympatric wildlife. Many wildlife populations in the Reserve are currently in decline (Ottichilo et al. 2000; Ogutu & Owen -Smith 2005; Ogutu et al. 2009; 2011 ; Chapters 2 and 3 of this dissertation ), and our findings reveal that spotted hyenas might be a useful tool in predicting these declines. Armed with this information, conservation biologists and wildlife management officials can study the behavior of spotted hyenas to understand how the effects of changing environments may manifest in other wildlife. Furthermore, 95 by recognizing the regions in which the movement behaviors of spotted hyenas are changing, it may be possible to identify and halt the disturbances that may be responsible . In the current research, a complex set of ecological and anthropogenic factors influenced the movements of spotted hyenas . The daily activity pattern of spotted hyenas in the Reserve is crepuscular and nocturnal, with no clear peaks throughout the night (Kolowski et al. 2007) . Our research adds to these earlier findings, but is the first to report that ambient temperature and amount of moonlight can also have an effect on the speed of movement of hyenas in the Reserve . The daily minimum and maximum temperatures near the Reserve have increased by 2.6 ¼C and 1.4 ¼C, respectively, from 1988 to 2011 (Chapter 2 of this dissertation), and daily temperatures are expected to continue to rise based on climate change models (Ritchie 2008) . In the current research, spotted hyenas traveled faster on nights with more moonlight, which is in contrast to previo us research (Cozzi et al. 2012) . This is also different from previous research on lions ( Panthera leo) , which found that lions move d slower in areas of high anthropogenic activity on nights w ith more moonlight (Oriol -Cotterill et al. 2015) . Future research should investigate how the amount of moonlight may be interacting with anthropogenic activities within and around the Reserve to affect hyena behavior, and how this in turn might affect the probability of conflict wi th people. More rainfall also led to faster movements by spotted hyenas . Previous researc h has documented slower movement rates by spotted hyenas during wet than dry seasons in other areas of Africa (Stratford & Stratford 2011) , but none has yet to document the effects of fine -scale rainfall patterns on the movements of spotted hyenas like we have here . 96 Using different methods, but consistent with previous research (Kolowski et al. 2007; Kolowski & Holekamp 2009) , we found that spotted hyenas living in regions of high anthropogenic disturbance consistently moved faster than did animals in more pristine areas. Interes tingly, the speed of movement also differed between inside and outside of Reserve boundaries, and was predicted by social rank. Previous research in the Reserve did not find a correlation between rates of hyena activity and social rank (Kolows ki et al. 2007) . Yet, the results from our study show animals of low social rank exhibiting faster movement in undisturbed regions than did high -ranking animals in the same areas. The space -use of spotted hyenas has previously been shown to vary with soci al rank; adult females of low social rank tend to range more widely when not limited by den-dwelling cubs, probably to avoid intra -specific competition while foraging (Boydston et al. 2003b) . However, the effects of anthropogenic activity inside the Talek region and outside of the Reserve may be affecting this relationship. Low-ranking hyenas in Talek moved at slower speeds than did high-ranking animals in the same areas , whereas we found the opposite relationship between low- and high-ranking hyenas in TMC (Figure 4.2) . Talek is t he region of the Reserve with the highest levels of anthropogenic disturbance ; thus, the influences of people on the speed of movement by hyenas appear to be stronger than the effects of certain natural intra -specific interactions. The tortuosity of the pa ths of spotted hyenas was predicted by the time of day, the amount of rainfall, and whether the animals were moving inside or outside of the Reserve (Table 4.2). Spotted hyenas inside of the Reserve had smaller turning angles than did those outside of the Reserve, indicating that hyenas were more likely to display 97 less straightforward movements in the community lands. The area immediately outside of the Reserve contains an expanding population of Maasai pastoralists (Lamprey & Reid 2004; Chapter 2 of this dissertation ). Many of these pastoral ists keep much of their wealth in livestock. In recent years, these communities have become reliant on grazing their livestock within the Reserve both day and night (Butt 2014 ; see also Chapter 2 of this dissertation ). Thus, the potential for conflict between large carnivores and people is now high both inside and outside of the Reserve. Although no previous res earch exists on effects of proximity to human settlements on the space -use or movement behaviors of spotted hyenas with GPS collars, research on lions has indicated variation in turning angles as a function of proximity to people. In areas where lions are more likely to come into conflict with people, turning angles are significantly lower than where the probability of conflict is lower (Oriol -Cotterill et al. 2015) . Using the same calculation for turning angle, but in contrast to Oriol -Cotterill et alÕs (2015) findings, we found that spotted hyenas exhibit more directed movements in areas inside of the Reserve boundaries than outside its borders. This may indicate that spotted hyenas in community lands display more erratic behaviors and are more disrupted than animals wit hin the Reserve. The fact that the rates of movement by spotted hyenas correlated with the local abundance of sympatric wildlife up to 26 weeks later could be helpful to wildlife managers. Although t his relationship may only assist in understanding short -term threats to wildlife in the environment, it can serve as another tool for wildlife managers to assess the status of this ecosystem. Monitoring herbivore and carnivore population trends over time is costly, and many years of data are often needed to appr opriately determine the status of wildlife populations. Furthermore, a general assessment of 98 wildlife abundance and diversity needs to occur well before it is possible to understand what ecological or anthropogenic influences may be causing any perceived c hanges. Research like ours represents at least an initial step in finding ways for indicator species not only to reflect current ecological conditions, but also to estimate future conditions . It remains to be seen whether the movements of hyenas will corre late with the abundance or diversity of sympatric wildlife in other areas of sub -Saharan Africa. Spotted hyenas are one of the few species in this region of Kenya in which population sizes appear to be increasing in areas of intensive anthropogenic distur bance (Chapter 3 of this dissertation). However, the changes in their movement behaviors and space -use to cope with these disturbances, like those documented here and in previous studies (Boydston et al. 2003a; Kolowski et al. 2007; Kolowski & Holekamp 2009) , are likely to be costly in the long -term and could ultimately lead to population declines. For example, we could start to see populations of spotted hy enas decline in the future if faster movements increase the probability of conflicts with people . Above all, when one of the most adaptive and plastic species in this environment is showing changes in its behavior, we can expect sympatric wildlife, which m ay be more limited in their abilities to adapt to changing environments, to be faring considerably worse. 99 CHAPTER 5: Temporal dynamics of the responses by African mammals to prescribed fire 100 Introduction Fire plays a significant role in structuring ec osystems in sub -Saharan Africa, both as a natural occurrence and as a management technique for increasing vegetation biomass, vegetation palatability, and biodiversity (Norton -Griffiths 1979; Archibald 2008; Driscoll et al. 2010) . Fire is an important tool commonly used by wildlife managers to maintain grassland vigor (Dublin 1986; Bond et al. 2005; Govender et al. 2006) . Fires clear nutrient -deficient biomass (Van de Vijver et al. 1999) , and in conjunction with appropriately timed precipitation, may enhance primary productivity and foraging opportunities for herbivores (Moe et al. 1990; Van de Vijver et al. 1999) . However, deliberate burning by wildlife managers can interfere with the maintenance of natural ecological processes if prescribed fire regimes do not mimic naturally occurring fires (Parr & Andersen 2006; Van Wilgen et al. 2007; Van Wilgen & Biggs 2011; Van Wilgen et al. 2011). Deliberate burning may also have long -term effects on arthropods, birds, and small mammals (Salvatori et al. 2001; Yarnell et al. 2007; Little et al. 2013) , making it particularly important t o understand the ecological consequences of burning . Even though the management practice of burning has been used for decades in sub -Saharan Africa, relatively little is known about the temporal dynamics of herbivore responses to prescribed burns (de Ronde et al. 2004; but see Sensenig et al. 2010; Driscoll et al. 2010). Wildlife managers have long used prescribed burning as a tool to ach ieve a variety of short - (< 1 year) and long -term ( & 1 year) goals . Although the primary short -term goal is to create new foraging opportunities for herbivores during dry seasons by increasing primary productivity (e.g., Hobbs & Spowart 1984; Moe et al. 1990; Vermeire 101 et al. 2004; Geldenhuys et al. 2004; Everson et al. 2004) , short -term effects of burns might include making carnivores easier for ecotourists to see and photograph . Long-term goals include improving habitat quality to increase the carrying capacity of an area for one or more herbivore species (e.g., Parrini & Owen -Smith 2010) , preventing grassland areas from becoming woodlands (e.g., Norton -Griffiths 1979; Dublin 1986) , and increasing carnivore population sizes by enhancing abundance of their prey (e.g., Dees et al. 2001) . Burning is also often undertaken by wildlife managers in hopes of inducing both short - and long-term increases in native plant and animal species diversity in a particular area (Hobbs & Spowart 1984; Moe et al. 1990; Vermeire et al. 2004; Geldenhuys et al. 2004; Ever son et al. 2004; Parrini & Owen -Smith 2010) , and by pastoralists as a means to control ectoparasites (Riginos et al. 2012) . New vegetative growth after fire is utilized by some herbivores, and may temporarily alter regional species abundance (Wilsey 1996; Dees et al. 2001; Klop & Prins 2008) . Previous research has found that large mammalian herbivores in sub -Saharan Africa respond to burning in 3 ways: 1) utilizing immediate post -fire conditions (< 30 days post -burn), 2) exploiting post -fire re -grow th of vegetation ( & 30 -365 days post -burn), and 3) using burned habitat after considerable re -growth has occurred (> 365 days post -burn; reviewed by de Ronde et al. (2004) . Although many wildlife species use burned areas at some point during vegetation re -establishment, not all species respond similarly (de Ronde et al. 2004) . In some regions of Africa interspecific variation in post -fire utilization by herbivores can be explained in part by body size; that is, a negative allometric relationship exists between body mass and tendency to use burned sites (Wilsey 1996; Sensenig et al. 2010) . However this relationship varies among 102 geographic region s of sub -Saharan Africa (Tomor & Owen -Smith 2002; Klop et al. 2007), and is further complicated by interspecific variation in mode of digestion, with ruminants showing stronger preferences for post -burn vegetation than non -ruminants (Sensenig et al. 2010) . Most research to date has focused on responses of African herbivores to prescribed burning, and virtually nothing is currently known about responses by African carnivores (but see Eby et al. 2013) . Burned areas should attract carnivores if prey are more common and more readily accessible t here than in unburned areas . However, the clearing of vegetation may also decrease capture success of ambush -style hunters that require adequate cover to capture prey . In North America, Florida panthers ( Puma concolor coryi ) utilized burned areas for up to 1 year after burning, presumably because of increased prey abundance in burned regions (Dees et al. 2001) . Based on their results, Dees et al. (2001) suggested that burned areas may support larger panther populations, and thus facilitate their conservation. Here we monitored wildlife use of 4 areas in the Mara -Serengeti ecosystem, both before and after burning, to document responses of indigenous mammals to prescribed fire . Our objectives were to determine which species responded positively to burning, and to quantify the duration of thos e responses . Specifically, we were interested in determining which of 8 herbivore species were more likely to be seen on burned than unburned patches, which of these species were seen on burned patches in higher densities than on unburned patches, and for those that were seen in higher densities, how long these effects lasted. Additionally, we wanted to quantify how prescribed burning might enhance visibility of small and large carnivores to ecotourists in this 103 ecosystem. We used line -transect sampling in a repeated measures experimental design and generalized linear models to quantify wildlife response to burning while simultaneously accounting for annual patterns of precipitation, season, and movements of migratory herbivores. Wildlife managers and conserv ationists can use the results from our study to improve the efficacy of prescribed fire regimes in grassland ecosystems in sub -Saharan Africa. Study area Our study was conducted from 2008 to 2011 in the Mara Triangle region of the Maasai Mara National Res erve (henceforth, the Reserve), Kenya, at 1¡ 25' 41.20"S, 34¡ 57' 35.14"E ( Figure 5 .1). The Mara Triangle, which is managed by The Mara Conservancy, encompasses approximately 500 km 2, and is bounded by the Siria Escarpment, the Mara River, and the internat ional border with Tanzania ( Figure 5 .1). The Mara Triangle is primarily open, rolling grassland savannah interspersed with bush -covered inselbergs, and riparian vegetation along seasonal and permanent watercourses . Daily minimum and maximum temperatures av eraged 15.5 ¡C ± 0.4 ¡C (SE) and 29.5 ¡C ± 0.5 ¡C (SE), respectively, with little seasonal variation . Precipitation varies seasonally, with most falling during long (March -May) and short (November -December) rainy periods each year, resulting in significant seasonal variation in local biomass and nutrient content of grasses (Boutton et al. 1988a; 1988b) . Dominant grasses available to herbivores included Themeda triandra during the dry seasons, and Digitaria macroblephara, Hyparrhenia filipendula, Pennisetum mezianum, and Setaria sphacelata after the long rains (Sinclair 1979; Dublin 1986) . The standing biomass, 104 comprised mainly of T. triandra during the dry seasons, serves as fuel for fires (Anderson & Talbot 1965; McNaughton & Banyikwa 1995; Everson et al. 2004) . The Reserve supports large herbivore and carnivore populations year -round, and also supports large herds of migratory wildebeest ( Connochaetes taurinus ) and zebra (Equus quagga ) from both the Serengeti to the south and the Loita region to the northeast ; these herds are usually present during the dry season from July to October each year (Bell 1971; Maddock 1979; Sinclair & Norton -Griffiths 1979; Stelfox et al. 1986). Migratory populations of ThomsonÕs gazelle (Eudorcas thomsoni) that reside north of the Reserve also move into the Reserve durin g the dry se ason (Stelfox et al. 1986). The large migratory herbivores (i.e. wildebeest, zebra) reduce standing grass biomass, and this change in veg etation also affects local gazelle distributions and abundances (Bell 1971; Maddock 1979) . For example, in the Sereng eti National Park, ThomsonÕs gazelle follow localized movements of wildebeest and zebra, capitalizing on new vegetative growth that is stimulated by removal of the standing biomass (Bell 1971). 105 Figure 5 .1. Herbivore and carnivore survey transect locations in the Ma asai Mara National Reserve, southwestern Kenya, 2008 -2011. The four burn transects monitored in the Reserve are shown with the dates on which burns were initiated and the sample size representing the total number of times each transect was sampled . The dark triangle indicates the location of our weather station. 106 Methods Line-transect sampling From 2 August 2008 through 20 July 2011, we sampled abundance of common savannah mammals along 4 permanent transects ( Figure 5 .1), each 5 -7 km long, before and after a single burn was initiated in each region . Burns took place on 16 June 2008, 10 January 2009, 11 July 2009, and 27 Jun e 2010, and ranged from 29.8 to 49.8 km 2 in total area . Each transect ( Figure 5 .1) was surveyed twice per month, once during the first half of the month (days 1 -15), and once during the second half (days 16 -31). Observers (1 or 2) sampled mammal abundance between 0700 and 1000 hours by driving transects at 15 -20 km/hr and recording all mammals within 100 m of the transect centerline, creating a 200 m -wide detection area. During a subset of surveys taken June -August 2011 (n = 55), we used a rangefinder and d istance sampling to estimate detection functions for the herbivores monitored here with the program ÔunmarkedÕ in R (Buckland et al. 2001, Fiske and Chandler 201 1). We did this in both burned and unburned areas , and estimated that detection probability was & 90 % for all sampled species within 100 m of the transect centerline under both conditions (Figure 5.2) . Although we were only counting animals within 100 m of the transect centerline, our inability to estimate detection probabilities repeatedly during our study to account for variation due to year, season or transect location may have resulted in incomplete detection, and therefore, underestimates of herbivore numbers on transects. However, this was most likely the case in unburned areas with thick vege tation, and particularly for the smaller animals counted. 107 Figure 5.2. Detection probabilities for the 8 herbivore species monitored in this study based on unpu blished data from D. S. Green. Detection distances were calculated using a rangefin der during June -August of 2011. Solid lines indicate the detection function for animals in burned areas, whereas dashed lines indicate the detection function for ani mals seen in non -burned areas. 01002003004005000.00.20.40.60.81.0zebra Distance (m) Detection probability 01002003004005000.00.20.40.60.81.0warthog Distance (m) Detection probability 01002003004005000.00.20.40.60.81.0Thomson's gazelle Distance (m) Detection probability 01002003004005000.00.20.40.60.81.0Grant's gazelle Distance (m) Detection probability 01002003004005000.00.20.40.60.81.0impalaDistance (m) Detection probability 01002003004005000.00.20.40.60.81.0topi Distance (m) Detection probability 01002003004005000.00.20.40.60.81.0hartebeest Distance (m) Detection probability 01002003004005000.00.20.40.60.81.0wildebeestDistance (m) Detection probability 108 Figure 5.2 contÕd. Figures with only one line indicate no differences between detection distances as a function of burned status. The vertical line indicates the maximum distance for which animals were recorded in the study, and the horizontal line indicates the 90 % detection probabilities in burned and unburned areas. The herbivore species recorded on each transect included zebra, warthog (Phacochoerus africanus ), ThomsonÕs gazelle, GrantÕs gazelle ( Nanger granti) , impala (Aepyceros melampus) , topi (Damaliscus lunatus jimela), CokeÕs hartebeest (Alcelaphus buselaphus cokei ), and wildebeest . For each individual herbivore detected along any transect, we categorized it as present in either a burned or unburned area to generate animal densities (count per km 2) for each species . Animal density was used because transect lengths v aried . Along some portions of transects, a mosaic of burned and unburned patches occurred . Such areas were only categorized as burned when > 50 % of the detection area was burned . Small carnivores recorded on transects included black -backed jackals ( Canis mesomelas ), side -striped jackals ( Canis adustus ), bat-eared foxes ( Otocyon megalotis ), and banded mongoose ( Mungos mungo ). Large carnivore species recorded were African lions ( Panthera leo ), spotted hyenas ( Crocuta crocuta ), leopards ( Panthera pardus ), and cheetahs ( Acinonyx jubatus ). Statistical analyses of herbivore data We quantified herbivore selection of burned areas using generalized linear modeling with the glm package in R (R Core Team 2015) . The model estimated herbivore density for each species based on season, burn status, precipitation, and 109 transect identity . Every transect survey in each 2 -week interval was the statistical replicate for analyses, with one replicate per transect for densitie s of animals found in both burned and unburned areas . Species for which we found that burn status was a significant parameter were considered to be selecting burned over unburned areas (ANOVA F test; P < 0.05). Burn status was coded as a binary variable ( burned or unburned) . We accounted for seasonal variation in vegetation, the annual movements of wildebeest, zebra, and ThomsonÕs gazelle, and the timing of burns by coding each density with a categorical variable (season) inclusive of the month during whic h each transect was sampled; November-February was considered the early wet season, March -June was considered the late wet season, and July -October was considered the dry season. July -October was also considered to be the period during which migratory herb ivores were present in our study area . Due to large effects of seasonality on herbivore densities in the Reserve, we investigated impacts of burning on herbivores # 365 Days Since Burn (DSB) such that each season was represented only once in the data set for each transect . We obtained daily pre cipitation from a local weather station ( Figure 5 .1) and summarized precipitation amounts (in mm) for the 2 -week period corresponding to each survey, in addition to coding a quadratic term to investigate non -linear effects of rainfall (precipitation 2). To account for potential effects of spatial autocorrelation, or the impact of burn location on herbivore density, transect identity (ID) was included as a fixed effect . We also included interaction terms in our models between total precipitation and burn stat us, and between burn status and season, to account for variation in local vegetation and presence or absence of migratory herbivores . Animal densities were 110 loge(density + 1) transformed to achieve normality . Hence, our global model for each herbivore speci es was: loge(density + 1) ~ 1 + season + burn status + transect + precipitation + precipitation 2 + (burn status ' precipitation) + (burn status ' season) + (burn status ' precipitation 2) We determined the significance of these factors with an ANOVA F-test of the glm global model. A type II error structure was utilized because precipitation is a large driver of seasonality in our study system, and this ensured that each main effect was fit against all other terms when evaluating significance. We evaluate d the global model for each herbivore species separately and used influence plots of residuals in R to ensure fit and identify any overly influential data points . Residuals were normally distributed, and no outliers were detected. To determine how long ani mals occurred in greater densities on burned than unburned sites, we compared density estimates between burned and unburned sites along transects during several 30 -day intervals following initiation of each burn . DSB = 0 on the date of the burning event . We ran an analysis of variance (ANOVA) across 30 day blocks to determine which species yielded significant F ratios (P < 0.05; Crawley 2007), before performing multiple linear comparisons on these groupings using a Tukey adjustment to discern in which 30 day bl ocks statistically different animal density estimates occurred between burned and unburned areas ( P < 0.05). 111 Impact of prescribed burning on carnivore visibility Carnivores were observed too infrequently to use a generalized linear modeling framework, so we tested the duration of carnivore responses to burning by using a PearsonÕs chi -squared test of independence with YatesÕ correction . We recorded all carnivores seen on transects, but whether they were seen in a burned or unburned patch was not noted due to sampling constraints; therefore carnivore numbers were only compared between pre - and post -burn . We compared the numbers of carnivores sighted during four time intervals: preburn, 1 -120 DSB, 121 -365 DSB, and 366 -748 DSB . Data for small and large carniv ores were analyzed separately. Estimating vegetation recovery with the Normalized Difference Vegetation Index (NDVI) The Normalized Difference Vegetation Index (NDVI) is a reliable indicator of the amount and quality of biomass in a particular region (reviewed in Pettor elli et al. 2005). In the current study, monthly averages of NDVI were obtained from the National Oceanic and Atmospheric Administration (NOAA) from the Moderate Resolution Imaging Spectroradiometer (MODIS) at a resolution of ~ 5 x 5 km . To investigate th e rate of vegetation recovery and its relationship to herbivore numbers and carnivore visibility on burn transects, we extracted NDVI values for the grid cells in which > 60 % of the transect was located for the burns that occurred 11 July 2009 and 27 June 2010. We then divided these values by 10,000 to restore them to a -1 to 1 range. Using NDVI, we estimated when vegetation biomass returned to pre -burn levels, how herbivore numbers varied over time in relation to vegetation recovery, and determined whethe r 112 variation in visibility of carnivores was spurious or instead corresponded to the clearing of local vegetation. Results Four different transects were surveyed a minimum of 28 times each (see legend for Fig 1), resulting in a total of 145 counts of large mammals on all transects between 2 August 2008 and 20 July 2011, documenting effects of 4 prescribed burns ( Figure 5 .1). Data were collected up to 136 days before burns, and up to 748 days after burns occurred . Herbivores were seen during 73.1 % of the su rveys and small and large carnivores were observed during 13.8 % and 11.0 %, respectively. Effects of season, transect location, and precipitation on herbivore densities As expected, season was a strong predictor of density for zebra ( F2,168 = 18.224, P # 0.001), ThomsonÕs gazelle ( F2,168 = 5.217, P = 0.006), and wildebeest ( F2,168 = 55.859, P # 0.001). Interestingly, it was also a strong predictor for warthog ( F2,168 = 3.532, P = 0.031) and impala ( F2,168 = 3.181, P = 0.0441), but not for other species ( P > 0.05). Transect ID was an important predictor for densities of GrantÕs gazelle ( F3,168 = 3.122, P = 0.027), impala ( F3,168 = 10.492, P # 0.001), and topi ( F3,168 = 5.926, P # 0.001), but not for any other species ( P > 0.05) . The fixed effects of total p recipitation and precipitation 2 were significant only for impala ( F1,168 = 4.964, P = 0.027 and F1,168 = 4.956, P = 0.027) . Results from each global model are presented in Table 5.1. 113 Table 5.1. The results from the ANOVA F-test and probabilities derived from generalized linear models that predicted herbivore densities based on ecological factors for 8 herbivore species in the Mara -Serengeti ecosystem, southwest Kenya, 2008 -2011. Bolded P values indicate statistically significant effects. Denominator df = 168 in all global models. Species F df p-value zebra season 18.224 2 ! 0.001 status 19.568 1 ! 0.001 transect 2.072 3 0.106 precipitation 0.046 1 0.831 precipitation 2 0.005 1 0.946 status:precipitation 0.010 1 0.920 status:season 0.251 2 0.779 status:precipitation 2 0.057 1 0.812 warthog season 3.532 2 0.031 status 34.563 1 ! 0.001 transect 1.775 3 0.154 precipitation 0.990 1 0.321 precipitation 2 1.708 1 0.193 status:precipitation 0.259 1 0.611 status:season 0.312 2 0.733 status:prec ipitation2 0.207 1 0.650 Thomson's gazelle season 5.217 2 0.006 status 83.472 1 ! 0.001 transect 1.392 3 0.247 precipitation 0.254 1 0.615 precipitation 2 0.143 1 0.706 status:precipitation 0.196 1 0.659 status:season 0.644 2 0.527 status:precip itation2 0.086 1 0.770 Grant's gazelle 114 Table 5.1 contÕd !season 0.253 2 0.7764 status 14.755 1 ! 0.001 transect 3.122 3 0.027 precipitation 0.813 1 0.369 precipitation 2 1.469 1 0.227 status:precipitation 0.236 1 0.628 status:season 0.160 2 0.852 status:precipitation 2 0.398 1 0.529 impala season 3.181 2 0.044 status 3.275 1 0.072 transect 10.492 3 ! 0.001 precipitation 4.964 1 0.027 precipitation 2 4.956 1 0.027 status:precipitation 0.163 1 0.687 status:season 1.984 2 0.141 status:prec ipitation2 0.402 1 0.527 topi season 1.851 2 0.160 status 51.309 1 ! 0.001 transect 5.926 3 ! 0.001 precipitation 0.071 1 0.790 precipitation 2 0.208 1 0.649 status:precipitation 0.407 1 0.525 status:season 3.146 2 0.046 status:precipitation 2 1.099 1 0.296 hartebeest season 0.223 2 0.800 status 0.318 1 0.574 transect 1.149 3 0.331 precipitation 1.076 1 0.301 precipitation 2 0.756 1 0.386 status:precipitation 0.005 1 0.944 status:season 1.142 2 0.322 status:precipitation 2 0.109 1 0.742 wildebeest 115 Table 5.1 contÕd !season 55.859 2 ! 0.001 status 2.944 1 0.088 transect 0.911 3 0.437 precipitation 2.279 1 0.133 precipitation 2 1.245 1 0.266 status:precipitation 0.053 1 0.818 status:season 1.196 2 0.305 status:precipitation 2 0.077 1 0.781 Effects of burning on herbivores Wildlife response to burning was species -specific, with zebra, warthog, ThomsonÕs gazelle, GrantÕs gazelle, and topi occurring in larger numbers in burned than unburned areas ( P < 0.05; Table 5.2). Densities of impala and wildebeest showed trends towards selecti ng burned areas ( P = 0.072 and P = 0.088, respectively ), but hartebeest showed no response to burn status ( P > 0.10) . The interaction between season and burned status was significant only for topi ( F2,168 = 3.146, P = 0.046). The interaction between burned status and precipitation was not significant for any species (P > 0.05). 116 Table 5.2. Results from the ANOVA F-test and probabilities derived from generalized linear models that predicted herbivore densities based on burn status in the Mara -Serengeti ecos ystem, southwest Kenya, 2008 -2011. df = 1,168 in all comparisons. Species F P-value zebra 19.568 # 0.001 warthog 34.563 # 0.001 Thomson's gazelle 83.472 # 0.001 Grant's gazelle 14.755 # 0.001 impala 3.275 0.072 topi 51.309 # 0.001 hartebeest 0.318 0.574 wildebeest 2.944 0.088 Temporal dynamics of burn effects on herbivores Using only those s pecies for which burn status was a predictor of density, we next assessed the temporal dynamics of these responses . ANOVA F-tests indicated significant differences ( P < 0.05) between numbers of animals seen on burned areas compared to unburned areas of tra nsects over time for ThomsonÕs gazelle ( F11,156 = 2.044, P = 0.028) and topi ( F11,156 = 2.070, P = 0.026), but not for zebra ( F11,156 = 1.073, P = 0.387), GrantÕs gazelle ( F11,156 = 0.890, P = 0.552), warthog ( F11,156 = 0.696, P = 0.741), or wildebeest ( F11,156 = 0.730, P = 0.709). The direct effects of burning on ThomsonÕs gazelle lasted up to 120 DSB, and we observed a significant increase in topi densities on transects from 181 -240 DSB ( Figure 5.3 ). 117 Figure 5.3 . Density of ThomsonÕs gazelle and topi on burned and unburned areas, relative to the time of the fire, in the Ma asai Mara National Reserve . ÒPreburnÓ includes averaged data collected prior to the burn . Each subsequent interval on the x -axis represents a 30 -day block of time after the burn . Error bars represent 95% confidence intervals around the means . Two asterisks indicate significant differences between densities on burned vs. unburned areas at P # 0.001; one asterisk indicates significant differences at P < 0.05. Impact of burning on carnivor e visibility Burning appeared to increase numbers of both small and large carnivores seen on transects ( Figure 5.4 ). Compared to pre -burn sightings, large carnivores were observed in higher numbers up to 120 days after burns ( !12 = 23.193, P # 0.001), but 118 not during later time blocks (121 -365 DSB: !12= 3.698, P = 0.104 and 366 -748 DSB: !12 = 2.647, P = 0.186) . Small carnivores were observed more frequently after than before burns for up to 365 days after burning (1 -120 DSB: !12 = 15.510, P # 0.001 and 121-365 DSB: !12 = 8.254, P = 0.006), but not more than one year after burning events (366 -748 DSB; !12 = 1.0714, P = 0.575). 119 Figure 5.4 . Carnivore abundance in burned areas, relative to the time of the fire, in the Maasai Mara National Reserve . Two aster isks indicate significant differences from preburn abundances at P # 0.001; a single asterisk indicates significant differences from preburn abundances at P < 0.05 ( !2 test of independence with YatesÕ correction) . Error bars represent 95% confidence intervals around the means . Sample sizes for each grouping, representing the number of transects run , appear below the bars. (n = 12) (n = 28) (n = 57) (n = 48) Small Large Days since burn Mean number observed 0.00.51.01.52.0Preburn 1 - 120 121 - 365 366 - 748 ***** 120 Vegetation recovery with NDVI Reduction of local vegetation biomass following the burns was clearly visible using NDVI ( Figure 5.5 ). Levels of vegetation decreased by nearly half following the burns, rem ained at these levels for approximately 120 days, and returned to pre -burn levels by 180 -210 days after burning ( Figure 5.5 ). Figure 5.5 . The response of local vegetation to prescribed burning assessed using the Normalized Difference Vegetation Index (N DVI) . 0.00.20.40.60.81.0Days since burn NDVI Preburn 306090120150180210240270300330360 121 Figure 5.5 contÕd. The values represented here are the averaged monthly values for the 2 ~ 5 x 5 km grid cells in which > 60% of the burn transects fell for the prescribed burns taking place on 11 Jul 2009 and 27 Jun 2010 . Values were grouped by thei r Days Since Burn (DSB), and ÒPreburnÓ indicates the 30 -day interval before burning was initiated . Error bars indicate 95% confidence intervals around the means. Discussion Although prescribed burns in the current study increased local abundances of some herbivore species (zebra, warthog, ThomsonÕs gazelle, GrantÕs gazelle, topi), these effects were surprisingly short -lived and variable, with significant differences in abundance between burned and unburned patches found only for ThomsonÕs gazelle and topi. Our results indicate that herbivore response to burning is influenced by complex relationships among species, season, location, precipitation, and time since burn. For many years, wildlife managers have viewed prescribed burning in sub -Saharan African sav annahs as important for creating new foraging opportunities for wildlife during dry seasons (Norton -Griffiths 1979; Archibald 2008; Driscoll et al. 2010) . New vegetative growth stimulated by burning is utilized by certain wildlife (Van de Vijver et al. 1999; de Ronde et al. 2004; Sensenig et al. 2010) , and may contribute to population growth over the long -term . However, here we found that burn effects did not last more than four months for any species including ThomsonÕs gazelles, which exhibited stronger and more immediate responses than any other herbivore monitored. We found that zebra, warthog, ThomsonÕs gazelle, GrantÕs gazelle, and topi responded favorably to burning, as was also found in some earlier studies (Moe et al. 1990; Wilsey 122 1996; Gureja & Owen -Smith 2002; de Ronde et al. 2004) . Our results suggest that zebra, warthogs, GrantÕs gazelle, and topi may utilize burned areas that occur within their normal home ranges, but tha t these species do not congregate in large numbers on burned areas . Furthermore, because transect was also a strong predictor of density for topi abundance, the differences between burned and unburned patches 181 -240 DSB should be viewed with caution, alth ough others have also observed that topis select post -fire regrowth (Gureja & Owen -Smith 2002) . Overall, our data suggest that long-range movements to use burned habitats are probably uncommon among zebra, warthogs, GrantÕs gazelle, and topi. Numbers of ThomsonÕs gazelle in burned areas remained elevated for # 120 days ( Figure 5.3 ). This finding is consistent with that from Wilsey (1996) who also found that ThomsonÕs gazelle utilize freshly burned areas . The immediate response of ThomsonÕs gazelle to burning is likely caused by enhanced nutrient availability and reduced cellulose con tent in forage shortly after burns . Sensenig et al. (2010) found that nutrient content of vegetation increases within 30 days of the initial burn, but that enhanced nutrient content only persists for # 90 days . This change in vegetation associated with increa ses in ThomsonÕs gazelle numbers is consistent with our NDVI data; bulk forage decreased immediately after burns, allowing for the stimulation of new growth . Gazelles have highly tactile lips and narrow muzzles that allow for preferential feeding on isolat ed patches of protein -rich foliage, an adaptation that facilitates foraging on new growth in burned patches (Bell 1971; Maddock 1979; Estes 1991) . During the seasonal migration of wildebeest and zebra in the Serengeti ecosystem, ThomsonÕs gazelle follow wildebeest and zebra herds that have stimulated growth of protein rich, 123 cellulose -poor grass shoots by their grazing activity (Bell 1971; Maddock 1979) . Our data suggest that prescribed burning has a similar effect . However, once the biomass of vegetation returns to approximately half its pre -burn values, 120 days after burning, ThomsonÕs gazelles no longer appear to select burned over unburned areas, so the effects of burning on this species are short -lived. Topi occurred in significantly higher numbers in burned areas 181 -240 days after the fire, affirming their preference for post -fire regrowth (Gureja & Owen -Smith 2002; de Ronde et al. 2004). This also coincides with the interval after burns in which NDVI data show a return to pre -burn levels ( Figure 5.5 ). Transect identity and the interaction between burn status and season also influenced the presence of topi in burned areas (Table 5.1). In some parts of the Reserve topi are known to defend traditional territories, and in other areas they defend small territories on leks during the breeding sea son (Bro-J¿rge nsen & Durant 2003) ; the significant effects of transect identity and the interaction between burn status and season on topi numbers observed here may reflect variation in territory quality for topis and the seasonality of their lekking behavior in the Re serve . Previous research in South Africa found that topi selected post -fire growth around 4 weeks after burning (Gureja & Owen -Smith 2002) , indicat ing that those topi exploited burns earlier than did the topi observed in our study . In that earlier study, a small population of topi was investigated (n = 23, compared to 4675 in the Reserve, as described by Ogutu et al. 2011), and their patt erns of territory use were unknown in South Africa . Perhaps lower salience of territory quality for topis in southern Africa than the Reserve affects the timing of their use of burned areas, and allows them to respond faster to the availability of burned p atches. 124 Transect location also affected responses of impala and GrantÕs gazelle to prescribed burning . Unfortunately, however, GrantÕs gazelle and impala were seldom seen on transects, being present on only 6.8 % and 14.4 %, respectively, of all transects sampled . Our infrequent sightings of impala and GrantÕs gazelles on transects, together with limited information about their territorial behavior in the Reserve, make interpretation of these results difficult . Responses of impala to prescribed burning were also affected by precipitation, whereas responses of other herbivores were not. Season was a strong predictor of density for zebra, ThomsonÕs gazelle, and wildebeest, as was expected based on previous research and their known migratory patterns in the Res erve (Stelfox et al. 1986) . Interestingly, season was also a strong predictor for warthog and impala, indicating that these species might also benefi t more at some times of year than others from the modification of local vegetation . The high degree of seasonality in the Reserve may have driven ephemeral responses to burning by herbivores in the current study . Other savannah ecosystems that have less ex treme seasonality with respect to both rainfall and large movements of herbivores than in the current study may see greater effects of burning on herbivore populations than those documented here. Little information exists on how prescribed burning affects African carnivores or the implications of burning for carnivore management . Although results from the current study are preliminary, they provide a starting point for understanding how prescribed burning in sub -Saharan African grasslands can be expected to affect indigenous carnivores . In our study, prescribed burning appeared to increase numbers of both 125 small and large carnivores, but our results must be cautiously viewed because we were unable to account for detection probability . No carnivores were obser ved on transects before the burns occurred . It is reasonable to assume that carnivores might utilize burned areas as prey become locally abundant and accessible . However, it is also plausible that removal of tall grass by burning might simply improve detec tion probability . We were unable to distinguish between these two hypotheses. Our data on large carnivores are consistent with both of these hypotheses because increases in numbers of large carnivores coincided with enhanced abundance of ThomsonÕs gazelle s during the first 120 days after burning, and because the recovery of vegetation following the burn event was complete by 210 days after burning . A study by Eby et al. (2013) found African lions were more likely to be tracked during daylight hours in unbu rned areas, suggesting that they were avoiding burned habitat . This result is contrary to our findings, and further underscores the need for future research on the impacts of prescribed burning on carnivore populations . However, our data for small carnivor es are more difficult to interpret . If larger numbers of these carnivores were due to our enhanced ability to detect them, we would not expect to see numbers of small carnivores differing significantly from pre -burn numbers & 210 days post -burn, when vegetation biomass returned to pre -burn values . Our data show significant increases over preburn numbers of small carnivores to persist until one year after a burn . We emphasize that the numbers of small carnivores seen were small, so our result for this group should be interpreted with caution . We would expect the effects of burning on the prey of small carnivores to be more likely negative than positive . The likely costs of prescribed burning to small carnivores include at least temporary reductions in numbers 126 of small mammals, ground -nesting birds, insects, reptiles, and amphibians . Such prey animals are vulnerable to prescribed burning, and burns are known from earlier research to have negative direct and indirect effects on their current and future population sizes (Russell et al. 1999; Salvatori et al. 2001; Archibald et al. 2005; Yarnell et al. 2007; Little et al. 2013). Management implications Creating new foraging opportunities for herbivores during dry seasons is an important reason given by wildlife managers for prescribed burning in sub -Saharan Africa (Hobbs & Spowart 1984; Moe et al. 1990; V ermeire et al. 2004; Geldenhuys et al. 2004; Everson et al. 2004) . Although the impacts of burning on local and migratory herbivore populations were surprisingly short -lived in the current study, even slight improvements in forage quality and quantity in the region may assist these populations during the dry seasons . Moreover, even small increases in herbivore population sizes due to enhanced forage quality after burning might have important positive implications for endangered carnivore populations (e.g., cheetah, lions) that depend on these herbivores as a food source . However, it is also important to note that large resident herbivores, such as elephants and buffalos (not addressed in this study), may require considerable bulk forage year round, and burn ing large areas may therefore have long -term detrimental effects on them . Herbivore population sizes may not necessarily increase due to burning, because all that can be observed in short -term studies of burn effects, including ours, is the redistribution of herbivores across the landscape . We do not yet know enough about longer -term effects of burning on population dynamics of 127 African wildlife, so these should be studied further . Prescribed burns may also interfere with practices in place to preserve natur al ecology if burns are incorrectly timed, occur too frequently, or are inappropriately intense (Van Wilgen et al. 2007) . Because of the known and potential ecological costs of burning (Russell et al. 1999; Salvatori et al. 2001; Yarn ell et al. 2007; Little et al. 2013), we recommend that managers understand the limitations of prescribed burns, clearly weigh the benefits of burning against its potential costs, and formulate clear objectives before initiating a program of prescribed burning. 128 LITERATURE CITED 129 LITERATURE CITED Adler, P. B., D. A. Raff, and W. K. Lauenroth. 2001. The effect of grazing on the spatial heterogeneity of vegetation. Oecologia 128:465Ð479. Anderson, G. D., and L. M. Talbot. 1965. Soil factors affecting the distribution of the grassland types and their utilization by wild animals on the Serengeti Plains, Tanganyika. Journal of Ecology 53:33Ð56. Archibald, S. 2008. African grazing lawns - How fire, rainfall, and grazer numbers interact to affect grass community states. Journal of Wildlife Management 72:492Ð501. Archibald, S., W. J. Bond, W. D. Stock, and D. H. K. Fairbanks. 2005. Shaping the landscape: Fire -grazer interactions in an Afri can savanna. Ecological Applications 15:96Ð109. Arsenault, R., and N. Owen -Smith. 2002. Facilitation versus competition in grazing herbivore assemblages. Oikos 97:313Ð318. Augustine, D. J., K. E. Veblen, J. R. Goheen, C. Riginos, and T. P. Young. 2011. Pat hways for positive cattle Ðwildlife interactions in semiarid rangelands. Pages 55 Ð71 in N. J. Georgiadis, editor. Conserving Wildlife in African Landscapes Kenyas Ewaso Ecosystem. Washington, D.C. Bell, R. H. V. 1971. A grazing ecosystem in the Serengeti. S cientific American 224:86Ð93. Berger, K. M., E. M. Gese, and J. Berger. 2008. Indirect effects and traditional trophic cascades: a test involving wolves, coyotes, and pronghorn. Ecology 89:818Ð828. Beschta, R. L., and W. J. Ripple. 2009. Large predators an d trophic cascades in terrestrial ecosystems of the western United States. Biological Conservation 142:2401Ð2414. Elsevier Ltd. Bhattacharjee, S., V. Kumar, M. Chandrasekhar, M. Malviya, A. Ganswindt, K. Ramesh, K. Sankar, and G. Umapathy. 2015. Glucocorti coid stress responses of reintroduced tigers in relation to anthropogenic disturbance in Sariska Tiger Reserve in India. Plos One 10:e0127626. 130 Bhola, N., J. O. Ogutu, H. -P. Piepho, M. Y. Said, R. S. Reid, N. T. Hobbs, and H. Olff. 2012a. Comparative change s in density and demography of large herbivores in the Masai Mara Reserve and its surrounding human -dominated pastoral ranches in Kenya. Biodiversity and Conservation 21:1509Ð1530. Bhola, N., J. O. Ogutu, M. Y. Said, H. -P. Piepho, and H. Olff. 2012b. The d istribution of large herbivore hotspots in relation to environmental and anthropogenic correlates in the Mara region of Kenya. Journal of Animal Ecology 81:1268Ð1287. Bolker, B., H. Skaug, A. Magnusson, and A. Nielsen. 2012, January 2. Getting started with the glmmADMB package. R Foundation for Statistical Computing. Bond, W. J., F. I. Woodward, and G. F. Midgley. 2005. The global distribution of ecosystems in a world without fire. New Phytologist 165:525Ð538. Boutton, T. W., L. L. Tieszen, and S. K. Imbamb a. 1988a. Biomass dynamics in grassland vegetation in Kenya. African Journal of Ecology 26:89Ð101. Boutton, T. W., L. L. Tieszen, and S. K. Imbamba. 1988b. Seasonal changes in the nutrient content of East African grassland vegetation. African Journal of Ec ology 26:103Ð115. Boydston, E., K. Kapheim, H. E. Watts, M. Szykman, and K. E. Holekamp. 2003a. Altered behaviour in spotted hyenas associated with increased human activity. Animal Conservation 6:207Ð219. Boydston, E., K. Kapheim, M. Szykman, and K. E. Hol ekamp. 2003b. Individual variation in space use by female spotted hyenas. Journal of Mammalogy 84:1006Ð1018. Boydston, E., K. Kapheim, R. C. van Horn, L. Smale, and K. E. Holekamp. 2005. Sexually dimorphic patterns of space use throughout ontogeny in the s potted hyena (Crocuta crocuta). Journal of Zoology 267:271Ð281. Boydston, E., T. Morelli, and K. E. Holekamp. 2001. Sex differences in territorial behavior exhibited by the spotted hyena (Hyaenidae, Crocuta crocuta ). Ethology 107:369Ð385. Bro-J¿rgensen, J. , and S. M. Durant. 2003. Mating strategies of topi bulls: getting in the 131 centre of attention. Animal Behaviour 65:585Ð594. Buckland, S. T., D. R. Anderson, K. P. Burnham, J. L. Laake, D. L. Borchers, and L. Thomas. 2001. Introduction to Distance Sampling: estimating abundance of biological populations. Oxford University Press, New York, USA. Butchart, S. H. M. et al. 2010. Global biodiversity: Indicators of recent declines. Science 328:1164Ð1168. Butt, B. 2014. The political ecology of ÒincursionsÓ: Livest ock, protected areas and socio -ecological dynamics in the Mara region of Kenya. Africa 84:614Ð637. Butt, B., and M. D. Turner. 2012. Clarifying competition: the case of wildlife and pastoral livestock in East Africa. Pastoralism: Research, Policy and Pract ice 2:1Ð15. Pastoralism: Research, Policy and Practice. Calenge, C. 2006. The package ÒadehabitatÓ for the R software: A tool for the analysis of space and habitat use by animals. Ecological Modelling 197:516Ð519. Caro, T. M. 2010. Conservation by proxy. I sland Press, Washington, Covelo, London. Caro, T., and G. O'Doherty. 1999. On the use of surrogate species in conservation biology. Conservation Biology 13:805Ð814. Carter, N. H., B. K. Shrestha, J. B. Karki, N. M. B. Pradhan, and J. Liu. 2012. Coexistence between wildlife and humans at fine spatial scales. Proceedings of the National Academy of Sciences 109:15360Ð15365. Ceballos, G., P. R. Ehrlich, A. D. Barnosky, A. Garcia, R. M. Pringle, and T. M. Palmer. 2015. Accelerated modern human -induced species lo sses: Entering the sixth mass extinction. Science Advances 1:e1400253. Chapron, G. et al. 2014. Recovery of large carnivores in Europe's modern human -dominated landscapes. Science 346:1517Ð1519. Cooper, S. M. 1991. Optimal hunting group size: the need for lions to defend their kills against loss to spotted hyaenas. African Journal of Ecology 29:130Ð136. 132 Cooper, S. M., K. E. Holekamp, and L. Smale. 1999. A seasonal feast: long -term analysis of feeding behaviour in the spotted hyaena ( Crocuta crocuta ). Africa n Journal of Ecology 37:149Ð160. Cozzi, G., F. Broekhuis, J. W. McNutt, L. A. Turnbull, D. W. Macdonald, and B. Schmid. 2012. Fear of the dark or dinner by moonlight? Reduced temporal partitioning among Africa's large carnivores. Ecology 93:2590Ð2599. Craf t, M. E., K. Hampson, J. O. Ogutu, and S. M. Durant. 2015. Carnivore communities in the greater Serengeti ecosystem. Pages 419 Ð450 in A. R. E. Sinclair, K. L. Metzger, S. A. R. Mduma, and J. M. Fryxell, editors. Serengeti IV: Sustaining biodiversity in a c oupled human-natural system. University of Chicago Press, Chicago. Craigie, I. D., J. E. M. Baillie, A. Balmford, C. Carbone, B. Collen, R. E. Green, and J. M. Hutton. 2010. Large mammal population declines in Africa's protected areas. Biological Conservat ion 143:2221Ð2228. Elsevier Ltd. Crawley, M. J. 2007. Analysis of Variance. Pages 449 Ð488 in The R Book. John Wiley & Sons, West Sussex, England. Creel, S. et al. 2013a. Conserving large populations of lions - the argument for fences has holes. Ecology Let ters 16:1413Ðe3. Creel, S., and N. Creel. 1996. Limitation of African wild dogs by competition with larger carnivores. Conservation Biology. Creel, S., D. Christianson, and P. Schuette. 2013b. Glucocorticoid stress responses of lions in relationship to gro up composition, human land use, and proximity to people. Conservation Physiology 1:cot021. Creel, S., J. E. Fox, A. Hardy, J. Sands, B. Garrott, and R. O. Peterson. 2002. Snowmobile Activity and Glucocorticoid Stress Responses in Wolves and Elk. Conservati on Biology 16:809Ð814. Crooks, K. R., and M. E. Soul”. 1999. Mesopredator release and avifaunal extinctions in a fragmented system. Nature 400:563Ð566. de Ronde, C., W. Parr, C. Parr, B. Brockett, and C. Geldenhuys. 2004. Fire effects on 133 flora and fauna. P ages 60 Ð87 in J. Goldammer and C. Ronde, editors. Wildland Fire Management Handbook for Sub -Saharan Africa. Global Fire Management Centre/Oneworld Books. Dees, C. S., J. D. Clark, and F. T. Van Manen. 2001. Florida panther habitat use in response to prescr ibed fire. The Journal of Wildlife Management 65:141Ð147. Dirzo, R., H. S. Young, M. Galetti, G. Ceballos, N. J. B. Isaac, and B. Collen. 2014. Defaunation in the Anthropocene. Science 345:401Ð406. Dloniak, S. M. 2006. Masai Mara Predator Research Project; Masai Mara National Reserve, Kenya. Pages 1 Ð18. Dorresteijn, I., J. Hanspach, A. Kecsk”s, H. Latkov⁄, Z. Mezey, S. Sug⁄r, H. von Wehrden, and J. Fischer. 2014. Human -carnivore coexistence in a traditional rural landscape. Landscape Ecology 29:1145Ð1155. Driscoll, D. A. et al. 2010. Fire management for biodiversity conservation: Key research questions and our capacity to answer them. Biological Conservation 143:1928Ð1939. Elsevier Ltd. Dublin, H. T. 1986. Decline of the Mara woodlands: The role of fire and elephants. PhD Dissertation. University of British Columbia. East, M. L., and H. Hofer. 2001. Male spotted hyenas ( Crocuta crocuta ) queue for status in social groups dominated by females. Behavioral Ecology 12:558Ð568. Elliot, N. B., S. A. Cushman, D. W. M acdonald, and A. J. Loveridge. 2014. The devil is in the dispersers: predictions of landscape connectivity change with demography. Journal of Applied Ecology 51:1169Ð1178. Estes, J. A. et al. 2011. Trophic downgrading of planet Earth. Science 333:301Ð306. Estes, J. A., and J. F. Palmisano. 1974. Sea otters: Their role in structuring nearshore communities. Science 185:1058Ð1060. Estes, R. D. 1991. Gazelles and Their Allies; Tribe Antilopini. Pages 63 Ð90 in The Behavior Guide to African Mammals. The Universit y of California Press, Berkeley 134 and Los Angeles, USA. Everson, T. M., C. S. Everson, C. de Ronde, and W. Trollope. 2004. Regional fire management: Objectives, practices and prescribed burning application. in Wildland fire management handbook for sub -Sahara Africa. Global Fire Management Centre/Oneworld Books. Fiske, I. J., and R. B. Chandler. 2011. unmarked: An R package for fitting hierarchical models of wildlife occurrence and abundance. Journal of Statistical Software 43:1Ð23. Fleishman, E., and D. D. Mu rphy. 2009. A realistic assessment of the indicator potential of butterflies and other charismatic taxonomic groups. Conservation Biology 23:1109Ð1116. Fleishman, E., J. R. Thomson, R. Mac Nally, D. D. Murphy, and J. P. Fay. 2005. Using indicator species t o predict species richness of multiple taxonomic groups. Conservation Biology 19:1125Ð1137. Frank, L. G. 1986. Social organization of the spotted hyaena Crocuta crocuta. I. Demography. Animal Behaviour 34:1500Ð1509. Frank, L. G., and R. Woodroffe. 2001. Be haviour of carnivores in exploited and controlled populations. Pages 419 Ð441 in J. Gittleman, S. M. Funk, D. W. Macdonald, and R. K. Wayne, editors. Carnivore conservation. Cambridge University Press. Frank, L. G., K. E. Holekamp, and L. Smale. 1995. Domin ance, demography, and reproductive success of female spotted hyenas. Pages 364 Ð384 in A. R. E. Sinclair and P. Arcese, editors. Serengeti II: dynamics, management, and conservation of an ecosystem. University of Chicago Press, Chicago, Illinois. Frank, L. G., S. E. Glickman, and I. Powch. 1990. Sexual dimporhism in the spotted hyaena (Crocuta crocuta ). Journal of Zoology 221:308Ð313. Geldenhuys, C., B. W. Van Wilgen, W. J. Bond, C. Van de Vijver, and C. de Ronde. 2004. Fire effects on the maintenance of bio diversity, soil and nutrients. Pages 88 Ð113 in Wildland Fire Management Handbook for Sub -Saharan Africa. Global Fire Management Centre/Oneworld Books. 135 Gelman, A., and J. Hill. 2007. Data Analysis Using Regression and Multilevel/Hierarchical Models. Cambrid ge University Press, New York. Gonz⁄lez -Su⁄rez, M., and E. Revilla. 2013. Variability in life -history and ecological traits is a buffer against extinction in mammals. Ecology Letters 16:242Ð251. Govender, N., W. Trollope, and B. W. Van Wilgen. 2006. The ef fect of fire season, fire frequency, rainfall and management on fire intensity in savanna vegetation in South Africa. Journal of Applied Ecology 43:748Ð758. Gregory, R. D., A. van Strien, P. Vorisek, A. W. Gmelig Meyling, D. G. Noble, R. P. B. Foppen, and D. W. Gibbons. 2005. Developing indicators for European birds. Philosophical Transactions of the Royal Society B: Biological Sciences 360:269Ð288. Gureja, N., and N. Owen -Smith. 2002. Comparative use of burnt grassland by rare antelope species in a lowveld game ranch, South Africa. South African Journal of Wildlife Research 32:31Ð38. Hayward, M. W. 2006. Prey preferences of the spotted hyaena (Crocuta crocuta) and degree of dietary overlap with the lion (Panthera leo). Journal of Zoology 270:606Ð614. Hayward, M. W., and G. I. H. Kerley. 2005. Prey preferences of the lion ( Panthera leo ). Journal of Zoology 267:309Ð322. Hazzah, L., S. Dolrenry, L. Q. Naughton, C. T. T. Edwards, O. Mwebi, F. Kearney, and L. Frank. 2014. Efficacy of two lion conservation programs in Maasailand, Kenya. Conservation Biology 28:851Ð860. Hebblewhite, M., C. A. White, C. G. Nietvelt, J. A. McKenzie, T. E. Hurd, J. M. Fryxell, S. E. Bayle y, and P. C. Paquet. 2005. Human activity mediates a trophic cascade caused by wolves. Ecology 86:2135Ð2144. Henschel, J. R., and J. D. Skinner. 1987. Social relationships and dispersal patterns in a clan of spotted hyaenas Croctua crocuta in the Kruger Na tional Park. South African Journal of Wildlife Research 22:18Ð24. Henschel, J. R., and J. D. Skinner. 1991. Territorial behaviour by a clan of spotted hyaenas Crocuta crocuta . Ethology 88:223Ð235. 136 Hilty, J., and A. Merenlender. 2000. Faunal indicator taxa selection for monitoring ecosystem health. Biological Conservation 92:185Ð197. Hobbs, N. T., and R. A. Spowart. 1984. Effects of prescribed fire on nutrition of mountain sheep and mule deer during winter and spring. The Journal of Wildlife Management 48:551Ð560. Hofer, H., and M. East. 1993. The commuting system of Serengeti spotted hyaenas: How a predator copes with migratory prey: II. Intrusion pressure and commuters' space use. Animal Behaviour 46:559Ð574. Hofer, H., and M. L. East. 2003. Behavioral proc esses and costs of co -existence in female spotted hyenas: a life history perspective. Evolutionary Ecology 17:315Ð331. Holekamp, K. E., and L. Smale. 1993. Ontogeny of dominance in free -living spotted hyaenas: juvenile rank relations with other immature in dividuals. Animal Behaviour 46:451Ð466. Holekamp, K. E., and S. M. Dloniak. 2010. Chapter 6 - Intraspecific variation in the behavioral ecology of a tropical carnivore, the spotted hyena. Advances in the Study of Behavior 42:189Ð229. Holekamp, K. E., J. O. Ogutu, H. T. Dublin, L. G. Frank, and L. Smale. 1993. Fission of a spotted hyena clan: Consequences of prolonged female absenteeism and causes of female emigration. Ethology 93:285Ð299. Holekamp, K. E., L. Smale, and M. Szykman. 1996. Rank and reproductio n in the female spotted hyaena. Journal of Reproduction and Fertility 108:229Ð237. Holekamp, K. E., M. Szykman, E. Boydston, and L. Smale. 1999. Association of seasonal reproductive patterns with changing food availability in an equatorial carnivore, the s potted hyaena (Crocuta crocuta ). Journal of Reproduction and Fertility 116:87Ð93. Homewood, K., and W. A. Rodgers. 1988. Pastoralism, conservation and the overgrazing controversy. Pages 111 Ð128 in D. Anderson and R. Grove, editors. Conservation in Africa: people, policies and practice. Cambridge University Press, Cambridge. 137 Homewood, K., and W. A. Rodgers. 1991. Maasailand Ecology: pastoralist development and wildlife conservation in Ngorongoro, Tanzania. Cambridge University Press, Cambridge. Homewood, K., E. F. Lambin, E. Coast, A. Kariuki, J. Kivelia, M. Y. Said, S. Serneels, and M. Thompson. 2001. Long -term changes in Serengeti -Mara wildebeest and land cover: Pastoralism, population, or policies? Proceedings of the National Academy of Sciences 98:12544Ð12549. Hıner, O. P., B. Wachter, M. L. East, W. J. Streich, K. Wilhelm, T. Burke, and H. Hofer. 2007. Female mate-choice drives the evolution of male -biased dispersal in a social mammal. Nature 448:798Ð801. IUCN. 2015. The IUCN Red List of Threatened Specie s. Version 2015.1. . Downloaded on 03 July 2015. Johnson, C. N., J. L. Isaac, and D. O. Fisher. 2007. Rarity of a top predator triggers continent -wide collapse of mammal prey: dingoes and marsupials in Australia. Proceedings of the Royal Society B: Biological Sciences 274:341Ð346. Karanja, G. 2003. Tourism impacts inside Masai Mara National Reserve. Pages 5 Ð16 in M. Walpole, G. Karanja, N. Sitati, and N. Leader -Williams, editors. Wildlife and people: Conflict and conservation in Masa i Mara, Kenya. IIED, London. Kessler, M. et al. 2011. Cost -effectiveness of plant and animal biodiversity indicators in tropical forest and agroforest habitats. Journal of Applied Ecology 48:330Ð339. Kiage, L. M. 2013. Perspectives on the assumed causes of land degradation in the rangelands of Sub -Saharan Africa. Progress in Physical Geography 37:664Ð684. Kissui, B. M. 2008. Livestock predation by lions, leopards, spotted hyenas, and their vulnerability to retaliatory killing in the Maasai steppe, Tanzania. Animal Conservation 11:422Ð432. Klop, E., and H. H. T. Prins. 2008. Diversity and species composition of West African ungulate assemblages: effects of fire, climate and soil. Global Ecology and Biogeography 17:778Ð787. Klop, E., J. V. Goethem, and H. H. d e longh. 2007. Resource selection by grazing 138 herbivores on post -fire regrowth in a West African woodland savanna. Wildlife Research 34:77Ð83. Kolowski, J. M., and K. E. Holekamp. 2006. Spatial, temporal, and physical characteristics of livestock depredatio ns by large carnivores along a Kenyan reserve border. Biological Conservation 128:529Ð541. Kolowski, J. M., and K. E. Holekamp. 2009. Ecological and anthropogenic influences on space use by spotted hyaenas. Journal of Zoology 277:23Ð36. Kolowski, J. M., D. Katan, K. R. Theis, and K. E. Holekamp. 2007. Daily patterns of activity in the spotted hyena. Journal of Mammalogy 88:1017Ð1028. Kruuk, H. 1972. The spotted hyena: A study of predation and social behavior. University of Chicago Press, Chicago, IL. Lampre y, R. H., and R. S. Reid. 2004. Expansion of human settlement in Kenya's Maasai Mara: what future for pastoralism and wildlife? Journal of Biogeography 31:997Ð1032. Landres, P. B., J. Verner, and J. W. Thomas. 1988. Ecological uses of vertebrate indicator species: A critique. Conservation Biology 2:316Ð329. Little, I. T., P. A. R. Hockey, and R. Jansen. 2013. A burning issue: Fire overrides grazing as a disturbance driver for South African grassland bird and arthropod assemblage structure and diversity. Bio logical Conservation 158:258Ð270. Lyamuya, R. D., E. H. Masenga, R. D. Fyumagwa, and E. R¿skaft. 2014. Human Ðcarnivore conflict over livestock in the eastern part of the Serengeti ecosystem, with a particular focus on the African wild dog Lycaon pictus . Oryx 48:378Ð384. Maddock, L. 1979. The ÒMigrationÓ and Grazing Succession. Pages 104 Ð129 in A. R. E. Sinclair and M. Norton -Griffiths, editors. Serengeti: Dynamics of an Ecosystem. The University of Chicago Press, Chicago, USA. McGeoch, M. A., and S. L. Chow n. 1998. Scaling up the value of bioindicators. Trends in Ecology and Evolution 13:46Ð47. 139 McKee, J. K., P. W. Sciulli, C. David Fooce, and T. A. Waite. 2003. Forecasting global biodiversity threats associated with human population growth. Biological Conser vation 115:161Ð164. McNaughton, S. J. 1983. Serengeti grassland ecology: The role of composite environmental factors and contingency in community organization. Ecological Monographs 53:291Ð320. McNaughton, S. J., and F. F. Banyikwa. 1995. Plant communities and herbivory. Pages 49Ð70 in A. R. E. Sinclair and P. Arcese, editors. Serengeti II: Dynamics, Management, and Conservation of an Ecosystem. The University of Chicago Press, Chicago, USA. Mills, M. 1990a. Kalahari Hyenas: Comparative Behavioral Ecology o f Two Species. The Blackburn Press. Mills, M. G. L. 1990b. Kalahari hyenas: Comparative behavioral ecology of two species. Chapman & Hall. Moe, S. R., P. Wegge, and E. B. Kapela. 1990. The influence of man !made fires on large wild herbivores in Lake Burung i area in northern Tanzania. African Journal of Ecology 28:35Ð43. Mora, C., D. P. Tittensor, S. Adl, A. G. B. Simpson, and B. Worm. 2011. How many species are there on earth and in the ocean? PLoS Biology 9:e1001127. Newmark, W. D. 2008. Isolation of Afric an protected areas. Frontiers in Ecology and the Environment 6:321Ð328. Norton -Griffiths, M. 1979. The influence of grazing, browsing, and fire on the vegetation dynamics of the Serengeti. Pages 310 Ð352 in A. Sinclair and M. Norton -Griffiths, editors. Sere ngeti: Dynamics of an Ecosystem. The University of Chicago Press, Chicaco, USA. Norton -Griffiths, M., M. Y. Said, S. Serneels, D. S. Kaelo, M. Coughenour, R. H. Lamprey, D. M. Thompson, and R. S. Reid. 2008. Land use economics in the Mara area of the Seren geti ecosystem. Pages 379 Ð416 in A. R. E. Sinclair, C. Packer, S. A. R. Mduma, and J. M. Fryxell, editors. Serengeti III: Human Impacts on Ecosystem Dynamics. The University of Chicago Press, Chicago and London. 140 Noss, R. F. 1990. Indicators for monitoring biodiversity: A hierarchical approach. Conservation Biology 4:355Ð364. Ogada, D. L. 2014. The power of poison: pesticide poisoning of Africa's wildlife. Annals of the New York Academy of Sciences 1322:1Ð20. Ogutu, J. O., and H. T. Dublin. 2002. Demography of lions in relation to prey and habitat in the Maasai Mara National Reserve, Kenya. African Journal of Ecology 40:120Ð129. Ogutu, J. O., and N. Owen -Smith. 2005. Oscillations in large mammal populations: are they related to predation or rainfall? African Journal of Ecology 43:332Ð339. Ogutu, J. O., H. P. Piepho, H. T. Dublin, N. Bhola, and R. S. Reid. 2008. Rainfall influences on ungulate population abundance in the Mara !Serengeti ecosystem. Journal of Animal Ecology 77:814Ð829. Ogutu, J. O., H. P. Piepho, H. T. Dublin, N. Bhola, and R. S. Reid. 2009. Dynamics of Mara ÐSerengeti ungulates in relation to land use changes. Journal of Zoology 278:1Ð14. Ogutu, J. O., N. Bhola, and R. S. Reid. 2005. The effects of pastoralism and protection on the density and dis tribution of carnivores and their prey in the Mara ecosystem of Kenya. Journal of Zoology 265:281Ð293. Ogutu, J. O., N. Owen -Smith, H. P. Piepho, and M. Y. Said. 2011. Continuing wildlife population declines and range contraction in the Mara region of Keny a during 1977 -2009. Journal of Zoology 285:99Ð109. Oriol -Cotterill, A., D. W. Macdonald, M. Valeix, S. Ekwanga, and L. G. Frank. 2015. Animal Behaviour. Animal Behaviour 101:27Ð39. Elsevier Ltd. Ottichilo, W. K., J. De Leeuw, A. K. Skidmore, H. H. T. Prins , and M. Y. Said. 2000. Population trends of large non -migratory wild herbivores and livestock in the Masai Mara ecosystem, Kenya, between 1977 and 1997. African Journal of Ecology 38:202Ð216. Ottichilo, W. K., J. de Leeuw, and H. H. T. Prins. 2001. Popula tion trends of resident wildebeest [ Connochaetes taurinus hecki (Neumann)] and factors influencing them 141 in the Masai Mara ecosystem, Kenya. Biological Conservation 97:271Ð282. Packer, C. et al. 2013. Conserving large carnivores: dollars and fence. Ecology Letters 16:635Ð641. Pangle, W. M., and K. E. Holekamp. 2010. Lethal and nonlethal anthropogenic effects on spotted hyenas in the Masai Mara National Reserve. Journal of Mammalogy 91:154Ð164. Parr, C. L., and A. N. Andersen. 2006. Patch mosaic burning for b iodiversity conservation: A critique of the pyrodiversity paradigm. Conservation Biology 20:1610Ð1619. Parrini, F., and N. Owen -Smith. 2010. The importance of post -fire regrowth for sable antelope in a Southern African savanna. African Journal of Ecology 48:526Ð534. Pettorelli, N., J. O. Vik, A. Mysterud, J. -M. Gaillard, C. J. Tucker, and N. C. Stenseth. 2005. Using the satellite -derived NDVI to assess ecological responses to environmental change. Trends in Ecology and Evolution 20:503Ð510. P”riquet, S., H. Fritz, and E. Revilla. 2014. The Lion King and the Hyaena Queen: large carnivore interactions and coexistence. Biological Reviews:1 Ð19. Polasky, S., J. Schmitt, C. Costello, and L. Tajibaeva. 2008. Larger -Scale Influences on the Serengeti Ecosystem: Natio nal and International Policy, Economics, and Human Demography. Pages 347 Ð377 in A. R. E. Sincalir, C. Packer, S. A. R. Mduma, and J. M. Fryxell, editors. Serengeti III: Human Impacts on Ecosystem Dynamics. The University of Chicago Press, Chicago and Londo n. Prins, H. H. T. 1992. The pastoral road to extinction: Competition between widllife and traditional pastoralism in East Africa. Environmental conservation 19:117Ð123. Prins, H. H. T. 2000. Competition between wildlife and livestock in Africa. Pages 51 Ð80 in H. H. T. Prins, J. G. Grootenhuis, and T. T. Dolan, editors. Wildlife conservation by sustainable use. Kluwer Academic Publishers, Boston. R Core Team. 2015. R: A language and environment for statistical computing. R Foundation for Statistical Computi ng, Vienna, Austria. URL http://www.R -project.org/. 142 Rasmussen, G. S. A., and D. W. Macdonald. 2012. Masking of the zeitgeber: African wild dogs mitigate persecution by balancing time. Journal of Zoology 286:232Ð242. Reid, R. S. 2012. Savannas of our birth: people, wildlife, and change in East Africa. University of California Press. Reid, R. S. et al. 2003. People, Wildlife and Livestock in the Mara Ecosystem: the Mara Count 2002:1Ð143. Riggio, J. et al. 2013. The size of savannah Africa: A lion's (Panthera leo) view. Biodiversity and Conservation 22:17Ð35. Riginos, C., L. M. Porensky, K. E. Veblen, W. O. Odadi, R. L. Sensenig, D. Kimuyu, F. Keesing, M. L. Wilkerson, and T. P. Young. 2012. Lessons on the relationship between livestock husbandry and biodiversi ty from the Kenya Long -term Exclosure Experiment (KLEE). Pastoralism: Research, Policy and Practice 2:10. Pastoralism: Research, Policy and Practice. Ripple, W. J. et al. 2014. Status and Ecological Effects of the WorldÕs Largest Carnivores. Science 343:1241484. Ritchie, M. E. 2008. Global Environmental Changes and Their Impact on the Serengeti. Pages 183 Ð208 in A. R. E. Sinclair, C. Packer, S. A. R. Mduma, and J. M. Fryxell, editors. Serengeti III: Human Impacts on Ecosystem Dynamics. The University of Chicago Press, Chicago and London, Chicago and London. Runyoro, V., H. Hofer, E. B. Chausi, and P. Moehlman. 1995. Long -Term Trends in the Herbivore Populations of the Ngorongoro Crater, Tanzania. Pages 146 Ð168 in A. R. E. Sinclair and P. Arcese, editors. Ser engeti II: Dynamics, management, and conservation of an ecosystem. University of Chicago Press, Chicago. Russell, K. R., D. H. Van Lear, and D. C. Guynn Jr. 1999. Prescribed fire effects on herpetofauna: Review and management implications. Wildlife Society Bulletin 27:374Ð384. Salvatori, V., F. Egunyu, A. K. Skidmore, J. de Leeuw, and H. A. M. van Gils. 2001. The effects of fire and grazing pressure on vegetation cover and small mammal populations in the Maasai Mara National Reserve. African Journal of Ecol ogy 39:200Ð204. 143 Schuette, P., A. P. Wagner, M. E. Wagner, and S. Creel. 2013a. Occupancy patterns and niche partitioning within a diverse carnivore community exposed to anthropogenic pressures. Biological Conservation 158:301Ð312. Schuette, P., S. Creel, a nd D. Christianson. 2013b. Coexistence of African lions, livestock, and people in a landscape with variable land use and seasonal movements. Biological Conservation 157:148Ð154. Elsevier Ltd. Sensenig, R. L., M. W. Demment, and E. A. Laca. 2010. Allometric scaling predicts preferences for burned patches in a guild of East African grazers. Ecology 91:2898Ð2907. Serneels, S., and E. F. Lambin. 2001. Impact of land -use changes on the wildebeest migration in the northern part of the Serengeti ÐMara ecosystem. Jo urnal of Biogeography 28:391Ð407. Serneels, S., M. Y. Said, and E. F. Lambin. 2001. Land cover changes around a major east African wildlife reserve: the Mara Ecosystem (Kenya). International Journal of Remote Sensing 22:3397Ð3420. Sinclair, A. R. E. 1979. The Serengeti Environment. Pages 31 Ð45 in A. R. E. Sinclair and M. Norton -Griffiths, editors. Serengeti. The University of Chicago Press, Chicago, USA. Sinclair, A. R. E., and M. Norton -Griffiths. 1979. Serengeti: Dynamics of an Ecosystem. The University o f Chicago Press, Chicago, USA. Sinclair, A. R. E., and P. Arcese. 1995. Serengeti II: Dynamics, management, and conservation of an ecosystem. University of Chicago Press, Chicago. Smale, L., S. Nunes, and K. E. Holekamp. 1997. Sexually Dimorphic Dispersal in Mammals: Patterns, Causes, and Consequences. Pages 181 Ð250 in. Elsevier. Sogbohossou, E. A., H. H. De Iongh, B. Sinsin, G. R. de Snoo, and P. J. Funston. 2011. HumanÐcarnivore conflict around Pendjari Biosphere Reserve, northern Benin. Oryx 45:569Ð578. Stelfox, J. G., D. G. Peden, H. Epp, R. J. Hudson, S. W. Mbugua, J. L. Agatsiva, and C. L. Amuyunzu. 1986. Herbivore dynamics in southern Narok, Kenya. The Journal of 144 Wildlife Management 50:339Ð347. JSTOR. Stratford, K. J., and S. M. C. Stratford. 2011. Fi ne-scale movements and use of space by spotted hyaena ( Crocuta crocuta ) on Ongava Game Reserve, Namibia. African Journal of Ecology 49:343Ð352. Talbot, L., and P. Olindo. 1990. Amboseli and Maasai Mara, Kenya. The World Bank, Washington, D.C. Thompson, M., and K. Homewood. 2002. Entrepreneurs, elites and exclusion in Maasailand: Trends in wildlife conservation and pastoral development. Human Ecology 30:107Ð138. Thompson, M., S. Serneels, and E. F. Lambin. 2002. Land -use strategies in the Mara ecosystem Keny a: A spatial analysis linking socio -economic data with landscape variables. in S. J. Walsh and K. A. Crews -Meyer, editors. Remote sensing and GIS applications for linking people, place, and policy. Kluwer Academic, Boston. Tomor, B. M., and N. Owen -Smith. 2002. Comparative use of grass regrowth following burns by four ungulate species in the Nylsvley Nature Reserve, South Africa. African Journal of Ecology 40:201Ð204. Van de Vijver, C. A. D. M., P. Poot, and H. H. T. Prins. 1999. Causes of increased nutrien t concentrations in post -fire regrowth in an East African savanna. Plant and Soil 214:173Ð185. Van Dyke, F. G., R. H. Brocke, H. G. Shaw, B. B. Ackerman, T. P. Hemker, and F. G. Lindzey. 1986. Reactions of mountain lions to logging and human activity. The Journal of Wildlife Management 50:95Ð102. van Horn, R. C., T. L. McElhinny, and K. E. Holekamp. 2003. Age estimation and dispersal in the spotted hyena (Crocuta crocuta). Journal of Mammalogy 84:1019Ð1030. Van Meter, P. E., J. A. French, S. M. Dloniak, H. E. Watts, J. M. Kolowski, and K. E. Holekamp. 2009. Fecal glucocorticoids reflect socio -ecological and anthropogenic stressors in the lives of wild spotted hyenas. Hormones and behavior 55:329Ð337. Van Wilgen, B. W., and H. C. Biggs. 2011. A critical asses sment of adaptive ecosystem 145 management in a large savanna protected area in South Africa. Biological Conservation 144:1179Ð1187. Elsevier Ltd. Van Wilgen, B. W., N. Govender, and H. C. Biggs. 2007. The contribution of fire research to fire management: a cr itical review of a long -term experiment in the Kruger National Park, South Africa. International Journal of Wildland Fire 16:519Ð530. Van Wilgen, B. W., N. Govender, G. G. Forsyth, and T. Kraaij. 2011. Towards adaptive fire management for biodiversity cons ervation: Experience in South African National Parks. koedoe 53:96Ð104. Veblen, K. E. 2008. Season - and herbivore -dependent competition and facilitation in a semiarid savanna. Ecology 89:1532Ð1540. Vermeire, L. T., R. B. Mitchell, S. D. Fuhlendorf, and R. L. Gillen. 2004. Patch burning effects on grazing distribution. Journal of Range Management 57:248Ð252. Walpole, M. J., and N. Leader -Williams. 2001. Masai Mara tourism reveals partnership benefits. Nature 413:771. Walpole, M., G. Karanja, N. Sitati, and N . Leader-Williams, editors. 2003. Wildlife and People: Conflict and Conservationin Masai Mara, Kenya. IIED Wildlife and Development Series no 14:1 Ð69. Watts, H. E., and K. E. Holekamp. 2008. Interspecific competition influences reproduction in spotted hyen as. Journal of Zoology 276:402Ð410. Watts, H. E., and K. E. Holekamp. 2009. Ecological determinants of survival and reproduction in the spotted hyena. Journal of Mammalogy 90:461Ð471. Wilmers, C. C., J. A. Estes, M. Edwards, K. L. Laidre, and B. Konar. 201 2. Do trophic cascades affect the storage and flux of atmospheric carbon? An analysis of sea otters and kelp forests. Frontiers in Ecology and the Environment 10:409Ð415. Wilsey, B. J. 1996. Variation in use of green flushes following burns among African ungulate species: The importance of body size. African Journal of Ecology 34:32Ð38. 146 Wittemyer, G., P. Elsen, W. T. Bean, A. C. O. Burton, and J. S. Brashares. 2008. Accelerated human population growth at protected area edges. Science 321:123Ð126. Woodroffe, R. 2001. Strategies for carnivore conservation: lessons from contemporary extinctions. Pages 61 Ð92 in J. Gittleman, R. K. Wayne, D. W. Macdonald, and S. M. Funk, editors. Carnivore Conservation. Cambridge University Press. Woodroffe, R., and J. R. Ginsber g. 1998. Edge effects and the extinction of populations inside protected areas. Science 280:2126Ð2128. Woodroffe, R., and L. G. Frank. 2005. Lethal control of African lions (Panthera leo): local and regional population impacts. Animal Conservation 8:91Ð98. Yarnell, R. W., D. M. Scott, C. T. Chimimba, and D. J. Metcalfe. 2007. Untangling the roles of fire, grazing and rainfall on small mammal communities in grassland ecosystems. Oecologia 154:387Ð402. Yirga, G., H. H. De Iongh, H. Leirs, K. Gebrihiwot, J. De ckers, and H. Bauer. 2012. Adaptability of large carnivores to changing anthropogenic food sources: diet change of spotted hyena ( Crocuta crocuta ) during Christian fasting period in northern Ethiopia. Journal of Animal Ecology 81:1052Ð1055. Yirga, G., W. E rsino, H. H. De Iongh, H. Leirs, K. Gebrehiwot, J. Deckers, and H. Bauer. 2013. Spotted hyena (Crocuta crocuta) coexisting at high density with people in Wukro district, northern Ethiopia. Mammalian Biology 78:193Ð197. Elsevier GmbH. Zipkin, E. F., B. Gardner, A. T. Gilbert, A. F. OÕConnell Jr., J. A. Royle, and E. D. Silverman. 2010. Distribution patterns of wintering sea ducks in relation to the North Atlantic Oscillation and local environmental characteristics. Oecologia 163:893Ð902.