Ecological implications of food and predation risk for herbivores in the Serengeti Hopcraft, John Grant Charles

Transcription

1 University of Groningen Ecological implications of food and predation risk for herbivores in the Serengeti Hopcraft, John Grant Charles IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite from it. Please check the document version below. Document Version Publisher's PDF, also known as Version of record Publication date: 21 Link to publication in University of Groningen/UMCG research database Citation for published version (APA): Hopcraft, J. G. C. (21). Ecological implications of food and predation risk for herbivores in the Serengeti. Groningen: s.n. Copyright Other than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons). Takedown policy If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim. Downloaded from the University of Groningen/UMCG research database (Pure): For technical reasons the number of authors shown on this cover page is limited to 1 maximum. Download date: 93219

2 Chapter 7 Serengeti wildebeest and zebra migrations are affected differently by food resources and predation risks. Authors: J. Grant C. Hopcraft, Juan M. Morales, Hawthorn L. Beyer, Daniel T. Haydon, Markus Borner, Anthony R.E. Sinclair, and Han Olff. Chapter 7 Reference: Formatted as a report in preparation for submission to Ecology. 143

3 Section II: Balancing Resources and Risk. ABSTRACT Migrations by terrestrial herbivores occur across predictable gradients such as rainfall or latitude, however the quality and quantity of food sought by migrants occurs at multiple spatial scales and is often ephemeral. Migrants must adopt search strategies which balance acquiring sufficient food while minimizing the risk of predation. Therefore, food quality, food abundance and predation might differentially affect the routes migrants choose. Data from 3 GPS radio collared wildebeest and zebra migrating seasonally in the SerengetiMara ecosystem illustrate that food and predation differentially affect the movement patterns of these cooccurring species. The daily step lengths and turn angles of wildebeest is determined primarily by the quality of grass rather than its abundance or the risk of predation. Conversely, zebra movement is best explained as accessing the most food of sufficient quality while avoiding predation. Both species tend to move directionally in proximity to water and do not linger, as these areas are frequented by predators. Furthermore, both species consistently move further each day when close to human habitation. The findings indicate wildebeest and zebra select different attributes of the same landscape, which increases our understanding of migratory behavior and highlights the importance of maintaining landscape heterogeneity and connectivity for the longterm conservation of migratory herbivores. INTRODUCTION Global declines of large mammal migrations due to human activity and a lack of longterm protection raises concerns about the persistence of landscapescale biological processes (Harris et al. 29). Since migrations rely on large contiguous habitats, the collapse of migratory systems worldwide is an indicator that many areas are succumbing to increasing human pressure (Bolger et al. 28). Migration enables herbivores to escape the limitations of local food supply as well as reduce their exposure to predation (Hebblewhite and Merrill 29) which releases them from regulation and enables the population to become superabundant (Fryxell et al. 1988). Because migratory herbivores such as wildebeest are so abundant in some ecosystems (Hopcraft et al. submitted) they have unusually large impacts. For example, the seasonal movement of millions of animals in the Serengeti affects ecosystem dynamics such as fire frequency and treegrass competition (Dublin et al. 199, Holdo et al. 29), biodiversity of grasses and animals (Anderson et al. 27b), as well as the socioeconomic status of local people (Sinclair and Arcese 1995, Sinclair et al. 28). Therefore, understanding how individual migrating animals choose to move on a daily basis is an important yet not well understood aspect of migrations (Alerstam 26, Schick et al. 28). Herbivore movement patterns typically orient around access to key resources (especially food), while minimizing the exposure to risk (especially predation or anthropogenic causes) (Fryxell et al. 28). For instance, the movement of elk (Cervus elaphus) in the Rocky Mountains is influenced by favorable microclimatic conditions and suitable food resources (Frair et al. 25), but also by proximity to risks such predation from wolves, or disturbance from roads and clearcut logging (Hebblewhite and Merrill 27, Frair et al. 28, Hebblewhite and Merrill 29). Similarly, the movement of Thomson s gazelle (Eudorcas thomsoni) in the Serengeti is closely related to periodic greening of the short grass sward which maximizes their energy gain (Fryxell et al. 24, Fryxell et al. 25). Furthermore, evidence suggests that herbivores behave differently when resources are distributed heterogeneously as opposed to homogenously over space (Cromsigt and Olff 26). For instance, elk tend to move longer distances and have a greater propensity for return movements (i.e. 18 turns) in heterogeneous habitats (Morales et al. 25). 144

4 Food and predation differentially affect wildebeest and zebra migrations. Establishing commonalities and variations in the movement patterns between different migratory species elucidates drivers of animal migration and clarifies which habitat features are critical for conservation. Until now, no studies have compared the movement patterns of two sympatric migratory species to ascertain if the same landscape variables influence the movement of both species equally. For instance, both wildebeest and zebra have similar migrations in the SerengetiMara ecosystem and yet these species are taxonomically unrelated (i.e. bovids versus equids) and with different digestive physiology. Ruminants, such as wildebeest, are more efficient at digesting moderate quality plant material than hindgut fermenters, such as zebra (Foose 1982, van Soest 1996). Zebra offset their lower digestive efficiency by processing greater quantities of forage faster which enables zebra to gain sufficient energy from lower quality grass (Bell 197, BenShahar and Coe 1992). Therefore, based on the digestive differences between these two unrelated species, wildebeest and zebra might make choices as to how to move through the same landscape based on different attributes of the habitat. The migratory behavior of wildebeest in the Serengeti enables the herds to maximize their access to high quality grass and thereby escape the local limitations of food supply (Wilmshurst et al. 1999). However, the overall population of wildebeest in the Serengeti is regulated by the abundance of dry season forage and not predation (Mduma et al. 1999). By comparison, previous studies suggest that the Serengeti zebra population is be limited by predation rates, especially on juvenile age classes, rather than by the overall food supply (Sinclair 1985, Grange et al. 24). Furthermore, because zebra are less nutritionally constrained by food quality than wildebeest, we might not expect zebra to migrate to the open nutrientrich short grass plains of the Serengeti, and yet they do (Skoog 1969)Maddock, 1979 #917}. Therefore, zebra might be choosing where and how to move during the migration based on factors related to predation, while wildebeest might make choices based on food quality. Previous research suggest zebra have a large spatial awareness (>3km) and move directionally probably in relation to grazing patches (Brooks and Harris 28), however movement in relation to predation risk has not been investigated until now. The objective of this study is to determine how food quality, food abundance, and predation risk influence the distance and direction migratory wildebeest choose to travel each day, and whether this differs from migratory zebra. We use individual based models of free ranging wildebeest and zebra to determine which landscape variables best describe the movement. We expect that the daily step lengths and turn angles of migratory wildebeest are determined by resource availability (such as grass nitrogen content and grass greenness) and by exposure to predation and anthropogenic risks (thick woody vegetation, tall grass, and proximity to human habitation). In contrast, zebra step lengths and turn angles are expected to be determined by predation parameters and the abundance of food, and not by food quality. Therefore, the daily movements of two sympatric migratory species might be determined by different attributes of the landscape which expands our understanding of the drivers of animal migrations. Chapter 7 MATERIAL AND METHODS Zebra and wildebeest movement were studied in the greater SerengetiMara ecosystem which lies on the border of Kenya and Tanzania in East Africa. The ecosystem extends from 1 3 to 3 3 South and 34 and East and is defined by the extent of the wildebeest migration (Figure 1). Semiarid savannas and grasslands dominate the south, with mixed Acacia and Commiphora woodlands spread over the central and northern areas which are interspersed with large treeless glades (Reed et al. 28, Sinclair et al. 28). The average annual rainfall increases from approximately 5mm in the southeast to over 12mm in the northwest, and falls primarily in the wet season (November to May). The ecosystem is described in detail by Sinclair et al. (28). 145

5 Section II: Balancing Resources and Risk. a b Kenya Kenya Tanzania Tanzania Figure 1. The greater SerengetiMara ecosystem lies between Kenya and Tanzania and coincides with a strong regional rainfall gradient. Wildebeest and zebra move seasonally between open grassed plains in the southeast to woodland and open savanna areas in the west and north. Field sampling points were distributed across the rainfall gradient and in different vegetation types. We used data from 17 female migratory wildebeest fitted with GPS radio collars between the years of 2 and 28 (except 22) and 13 female zebra from 25 to 28 (see online Appendix B for details on animal capture, handling, and GPS collars). Only daily GPS fixes at 18: hours were selected. 146 Statistical analysis The objective of this study is to understand how different environmental variables related to food abundance, quality and risks affect the movement patterns of zebra and wildebeest. We model daily movement as correlated random walks where parameters of the distributions of steps and turns were functions of environmental variables. We used the Weibull distribution for daily step lengths and the wrapped Cauchy distribution for turning angles between consecutive daily steps (Morales et al. 24). The Weibull is a nonnegative continuous distribution defined by a scale parameter α and a shape parameter β. The flexible nature of the Weibull distribution approximates the distribution of movement well because it not only describes the distribution of the distance moved under such forms of movement as correlated random walks, but it also approximates the distribution of the distance moved under simple diffusion (Morales et al. 24). Therefore, by describing the shape parameter α (which describes the mean step length in kilometers) and the scale parameter β (which describes the variance around the mean) the Weibull distinguishes between many different possible modes of movement (e.g. distinguishing between correlated consistent short daily steps, as opposed to random periodic long daily step lengths, as opposed to simple diffusion). The wrapped Cauchy is a circular distribution defined by the parameters ρ and μ. The parameter ρ takes values between zero and one and describes the concentration around the mean such that as ρ approaches 1, there is a greater concentration around the mean which suggests directional

6 Food and predation differentially affect wildebeest and zebra migrations. persistence in movement while ρ near indicates equal probability in all directions (i.e. random direction). The parameter μ describes the mean direction in radians and varies from to 2π (i.e. to 36 ) which is used to define the turn angle between consecutive steps. Therefore, directional persistence in movement can be distinguished from movement in random directions by estimating the parameters ρ and μ. Furthermore, the tendency for return movement towards previously occupied patches (i.e. turns of 18 ) as opposed forward movement towards new patches can also be evaluated. We use data on food resources and predation risk to estimate the mean step length (α), as well as the parameters ρ and μ that describe turn angles for the observed daily movement patterns of wildebeest and zebra. The basic model assumes that animal movement is a correlated random walk, such that the current turn angle is relative to the direction of the previous step (Zollner and Lima 1999). WinBUGS was used to build a Bayesian approach to a logistic regression where we ask what landscape variables best predict the observed movement parameters ρ, μ and α (see Appendix A). We only report variables where 75% of the posterior distribution is greater or less than zero for more than half the animals tested. For instance, a logistic regression model that included variables of food quality (such as grass nitrogen) but none of the variables that estimate predation risk (such as proximity to thick woody cover) would suggest that movement choices were determined by food quality, and not by predation. The best competing models were selected using the Deviance Information Criterion (DIC), a generalization of the AIC but applicable to Bayesian models using Markov chain Monte Carlo (MCMC) simulations. A comparison of zebra and wildebeest movement by broad habitat zones suggests the step lengths and turn angles are most different while animals are on the southern plains and most similar while animals migrate through the mixed woodlands of the Western Corridor, the central woodlands and the Masai Mara (Figure 2 and 3). Therefore, the parameters describing mean daily step lengths (α) and the direction of movement (ρ and μ) were estimated for both wildebeest and zebra on the southern plains separately from the mixed woodlands. Estimating resources and risks over the landscape GIS layers for seven predictor variables estimating food quality, food abundance and predation at a resolution of 1km2 were created and then extracted for each GPS location an animal was observed. Food quality was estimated from (1) the concentration of nitrogen in the grass, (2) the 16day mean NDVI value at the time of observation (i.e. the average greenness of the vegetation), and (3) the difference between the current 16day mean NDVI and the previous 16day mean NDVI values (positive values indicate drying, while negative values indicate greening). Grass nitrogen was measured at 148 sites (Figure 1) distributed in the different soil and vegetation types in Serengeti and across the rainfall gradient. Since the concentration of nitrogen in the grass is inversely correlated with the mean NDVI we regression kriged (Hengl et al. 27, Bivand et al. 28) the data from the 148 points with a 9year mean NDVI layer (2 29) to generate an accurate full grid representing the spatial distribution of grass nitrogen across the ecosystem (details provided in Appendix C). Grass biomass (4) is positively correlated with soil moisture and rainfall, and negatively correlated with grass quality (Breman and De Wit 1983, McNaughton et al. 1985, Olff et al. 22, Anderson et al. 27a). Therefore we used a topographic wetting index combined with the longterm average rainfall over a 46 year period to estimate the biomass of grass available to the migrants as opposed to the grass quality (see Appendix C). Landscape features such as dense thickets or water sources conceal predators or provide predictable locations where predators might encounter prey (Hebblewhite et al. 25, Hopcraft et al. 25, Balme et al. 27, Kauffman et al. 27, Valeix et al. 29b, Grace et al. in press). Therefore, we used (5) the distance to thick woody cover, and (6) the distance to permanent water sources to estimate the risk of predation. The mean horizontal woody cover 147 Chapter 7

7 Section II: Balancing Resources and Risk. available to stalking predators for each of 27 physiognomic landcover classes identified by Reed (Reed et al. 28) was determined at 1,882 points conducting along transects over the entire ecosystem (Figure 1). The only sources of permanent water are along the lower reaches of the largest rivers (see Appendix C). Exposure to human disturbance, such (9) as illegal hunting, was estimated by measuring the proximity to human settlements and scaled by the density of people. Large values indicate proximity to high density villages while small values are distant from low density villages (see Appendix C). We allowed for nonlinear relations by including a quadratic term for grass nitrogen, grass biomass, water, woody cover, and humans. Quadratic terms were not included in the final model if the model did not include the linear term. 148 RESULTS The longest daily step lengths (i.e. net daily displacement) for wildebeest and zebra occur on the southern plains (Figure 2 and 3), while the step lengths in Western Corridor, the central woodlands and the Masai Mara are all shorter. Wildebeest have the greatest propensity to move forward with few 18 turns except on the southern plains; the longest steps (i.e. > 12 km/ day) generally occur up to 45 left of right of a straight line (Figure 2). Wildebeest take the longest steps on the plains towards the east, south or west but not north. In the Western Corridor, wildebeest take the largest steps towards the south and east. By comparison, zebra frequently return towards areas they were on previous days (i.e. turn angles tend to approach either or 18 ). The longest step lengths (i.e. > 12 km/ day) rarely occur to the right or left (i.e. +9 and 9 ). The shortest step lengths for zebra occur in the Masai Mara where zebra seldom move more than 4 km / day. The results from the Markov chain Monte Carlo (MCMC) simulations of wildebeest and zebra movement (Table 1 and 2, respectively) suggest that zebra turn angles and step lengths are influenced more by predation and food abundance parameters than wildebeest. Wildebeest daily movement is determined by the quality of food (as estimated by NDVI and grass nitrogen) which affects zebra less. Both species respond similarly to water. Proximity to areas of high human density consistently results in longer step lengths for wildebeest and zebra regardless of the habitat they occupy. None of the landscape variables affect the direction of the turn (μ). The distance wildebeest travel each day on the plains is determined by the grass nitrogen concentrations and the proximity to water as described by positive quadratic terms (Table 1). Wildebeest step lengths are short at low grass nitrogen concentrations and also at very high grass nitrogen concentrations, but long when nitrogen levels are moderate. Similarly, wildebeest on the plains tend to have short daily step lengths near permanent water and very far from water, but longer step lengths at moderate distances. Grass biomass and proximity to thick woody vegetation do not affect the turn angles or step lengths of wildebeest on the plains. In the woodlands, wildebeest move directionally when grass nitrogen is low, or when grass biomass is low, or when they are near to thick woody cover or near to water (Table 1). The negative values for ρ implies there is greater concentration around the mean direction when the variable is low (i.e. directional movement), but less concentration around the mean direction when the variable is high (i.e. there is an equal chance of any direction being selected, such as in random movement). Furthermore, wildebeest move longer distances each day if the grass nitrogen is high, or when the area is greening, or if it is green already. When the biomass of grass is at moderate levels, wildebeest tend to move less each day, but will increase the distance they move when the biomass is either too low or too high (i.e. a negative quadratic relation). Thick woody vegetation which conceals stalking predators affects the turn angles of wildebeest but not the step lengths. However, when wildebeest are near water in the

8 Food and predation differentially affect wildebeest and zebra migrations. woodlands they tend to take longer steps than when they are distant from water (i.e. a negative α value). Zebra move more directionally on the plains when the grass biomass is very low, or when they are near thick woody vegetation, or near water (i.e. negative ρ values) (Table 2). The quality of the grass generally does not influence the direction or distance zebra move, except for NDVI (zebra move more when the area is green). The distance zebra move on the plains is however affected by the grass biomass; zebra tend to move further at moderate quantities of biomass than when the biomass is either too low or too high. Additionally, zebra move longer distances each day when they are near thick woody cover, or when they are very far from it (i.e. a negative quadratic for α). Zebra also tend to move further each day when they are distant from water. In the woodlands, zebra move more directionally when the grass nitrogen is low, or when the biomass of grass is low, or when they are close to water (negative values for ρ) (Table 2). Zebra move further each day when the areas are greener (larger NDVI) or have higher concentrations of grass nitrogen. Furthermore, zebra in the woodlands move long distances when biomass is very low or when it is very high, but short distances at moderate quantities of grass biomass (i.e. a negative quadratic affect). Zebra migrating through the woodlands also tend to move less each day when woody cover becomes very thick. DISCUSSION The most important finding from this study is that two cooccurring migratory species do not use the same cues to select their routes, even across the same landscapes. The daily movements of migratory wildebeest are made primarily to maximize their intake of high quality food. Conversely, sympatric zebra move so as to maximize their intake of the most food of sufficient quality while simultaneously minimizing their exposure to predator habitats. Wildebeest and zebra migrate to the southern short grass plains of the Serengeti for the wet season (Pennycuick 1975, Maddock 1979) because this area has an abundance of high quality grass which is available only for approximately 5 months of the year (Kreulen 1975, McNaughton 1985). During their time on the plains, wildebeest and zebra tend to move more each day than at any other point during the migration (Figure 2 and 3). Localized thunder showers and shallow soils result in fast greening and drying processes on the plains to which the grazers might be responding, and combined with the very high energy requirements of lactating females and rapid local depletion by large groups (Hopcraft et al. submitted), most likely accounts for the large daily movements during the wet season (i.e. greening and drying might be occurring faster than can be detected by the 16day NDVI composite images). Furthermore, the short grass, flat topography and lack of obstacles such as rivers or thick vegetation on the plains enables wildebeest and zebra to move relatively easily. Wildebeest on the plains tend to move longer distances each day as the concentration of nitrogen in the grass increases. However, when the grass becomes very nutrient rich wildebeest tend to settle and move less (Table 1). The optimal daily step length of wildebeest in relation to grass nitrogen suggests that individual wildebeest move so as to maximize their daily energy intake, which concurs with previous studies (Wilmshurst et al. 1999). Wildebeest movement on the plains is not related to predation or food abundance which differs from the behavior of zebra on the plains. The quality of grass on the plains does not generally affect the movement of zebra (with the exception of NDVI zebra move further when the grass is greener) (Table 2), which suggests that from a zebra s perspective the grass is equally nutritious anywhere on the plains. However, when zebra on the plains are close to thick woody vegetation they tend to move directionally and for long distances, which suggests they might become restless in areas with thick Chapter 7 149

9 Section II: Balancing Resources and Risk. Table 1. Landscape variables effecting the daily turn angles and step lengths of migrating wildebeest in the Serengeti. (+ indicates that at least 75% or more of the posterior is greater than for at least half of the animals. indicates that at least 75% or less of the posterior is less than for at least half of the animals. * indicates that at least 97.5% or more of the posterior is more than or less than for at least half the animals. n/a indicates variables that were not included in the final model) ρ describes the concentration around the mean turn angle for the wrapped Cauchy distribution and μ describes the direction of the turn. α is the scale parameter for a Weibull distribution which defines the mean distance of the step length. Plains Woodlands Turn Angle Step Length Turn Angle Step Length ρ μ α ρ μ α Food quality Nitrogen + + (Nitrogen)2 n/a NDVI +* dndvi Food Grass Biomass (Grass Biomass)2 n/a n/a + Predation Woody Cover (Woody Cover)2 n/a n/a Water + (Water)2 n/a n/a n/a woody cover because it potentially conceals predators such as lion (Hopcraft et al. 25). Furthermore, when the biomass of grass is very low on the plains, zebra tend to move directionally and do not return to previously occupied patches. In the highest biomass patches on the plains (which are still relatively low compared to the rest of the ecosystem (McNaughton 1985)), zebra tend to settle and move less. Zebra must eat large quantities of grass in order to extract sufficient energy due to the limitations of their hindgut digestive system (Bell 197, Foose 1982, van Soest 1996). Furthermore, evidence suggests that the Serengeti zebra popula Anthropogenic Human + + (Human)2 15

10 Food and predation differentially affect wildebeest and zebra migrations. Frequency Mixed Woodland (Masai Mara) Weibull parameters Scale = 3.65 (.17) Shape = 1.26 (.5) W Turn Angle 18 Bearing N S 9 1 % 1 1 % E WILDEBEEST Frequency Mixed Woodland (Central) 9 Weibull parameters Scale = 4.41 (.17) Shape = 1.3 (.3) W Turn Angle 18 Bearing N S 9 1 % 1 1 % E Daily step length (km) Daily step length (km) Frequency Mixed Woodland (Western Corridor) Weibull parameters Scale = 5.3 (.48) Shape = 1.4 (.8) W Turn Angle 18 Bearing N S 9 1 % 1 1 % E Legend Step length (km / day) >12 Frequency Plains (Southern Plains) 9 Weibull parameters Scale = 4.69 (.3) Shape =.99 (.4) W Turn Angle 18 Bearing N S 9 1 % 1 1 % E Chapter Daily step length (km) Daily step length (km) Figure 2. The daily movement behavior of wildebeest changes as they migrate. The longest daily step lengths occur on the plains which are the wet season range and the shortest steps occur in the northern dry season refuge of the Masai Mara. The largest steps (>12 km /day) generally occur between 45 to the left or right and wildebeest rarely turn 18 around towards recently occupied patches. 151

11 Section II: Balancing Resources and Risk. Table 2. Landscape variables effecting the daily turn angles and step lengths of migrating zebra in the Serengeti. (symbols are the same as for Table 1). Plains Woodlands Turn Angle Step Length Turn Angle Step Length ρ μ α ρ μ α Food quality Nitrogen + (Nitrogen)2 n/a n/a NDVI +* + dndvi Food Grass Biomass + * (Grass Biomass)2 +* Predation Woody Cover * (Woody Cover)2 n/a +* + n/a Water + (Water)2 n/a n/a Anthropogenic Human + + (Human)2 tion is most likely predator regulated (Sinclair and Norton Griffiths 1982, Grange et al. 24). Our findings illustrate that zebra move around the Serengeti plains so as to maximize their intake of abundant high quality grass, while minimizing their exposure to predation which concurs with previous research (Sinclair and Norton Griffiths 1982, Grange et al. 24, Brooks and Harris 28, Groom and Harris 21) and links population regulation to daily movement decisions. 152 The movement behavior of wildebeest and zebra changes substantially once the animals move into the woodland habitats towards their northern dry season refuge (Inglis 1976). In general, wildebeest movement in the woodlands continues to be directional with few turns beyond 45 to the left or right (Figure 2). Wildebeest tend to move even more directionally under low grass nitrogen conditions but for shorter distances (Table 1), which implies they do not return to recently grazed low nitrogen patches. In addition, decreasing grass quality in the woodlands (such as drying or dried grass) tends to cause wildebeest to move less each day, which was a trait previously observed but never conclusively explained (Inglis 1976). Conversely, as the grass greens or the nitrogen concentration increases, wildebeest move

12 Food and predation differentially affect wildebeest and zebra migrations. Frequency Mixed Woodland (Masai Mara) 9 Weibull parameters Scale = 2.66 (.25) Shape = 1.18 (.9) W Turn Angle 18 Bearing N S 9 1 % 1 1 % E ZEBRA Mixed Woodland (Central) Frequency Weibull parameters Scale = 4.1 (.9) Shape = 1.13 (.2) W Turn Angle 18 Bearing N S 9 1 % 1 1 % E Daily step length (km) Daily step length (km) Frequency Mixed Woodland (Western Corridor) Weibull parameters Scale = 3.93 (.18) Shape = 1.14 (.4) W Turn Angle 18 Bearing N S 9 1 % 1 1 % E Legend Step length (km / day) >12 Frequency Plains (Southern Plains) 9 Weibull parameters Scale = 5.19 (.13) Shape = 1.29 (.3) W Turn Angle 18 Bearing N S 9 1 % 1 1 % E Chapter Daily step length (km) Daily step length (km) Figure 3. Zebra tend to take longest steps between days on the southern plains and the shortest step in the northern dry season refuge of the Masai Mara. Zebra often return to previously occupied patches (i.e. turns of 18 ) and the largest steps (>12 km /day) generally when moving forward or returning (around or 18 ) and rarely to the left or right (9 or 9 ). 153

13 Section II: Balancing Resources and Risk. further each day, most likely in anticipation of the return migration to the southern plains at the beginning of the wet season. Wildebeest respond to very low grass biomass in the woodlands by moving long distances and in straight lines. They also tend to move long distances when the biomass is too high, presumably because excessive grass could potentially conceal predators (Packer et al. 25). Similarly, wildebeest respond to thick woody cover by moving directionally with few return movements, presumably as an antipredator response however they do not move further than normal. In general, wildebeest are most nutritionally constrained during the dry season (Mduma et al. 1999). The results indicate that wildebeest move during this time primarily so as to maximize their access to the most high quality grass. Unlike wildebeest, zebra show a greater propensity for return movement towards previously occupied patches (i.e. 18 turns, Figure 3), however they tend not to return to patches in the woodlands when grass biomass is low (i.e. directional movement as shown in Table 2). In addition, low grass biomass induces zebra to move further each day, presumably in search of food. At moderate quantities of grass biomass in the woodlands zebra move less suggesting a more encamped phase (Morales et al. 24, Fryxell et al. 28), however they move large distances again when grass biomass is very high most likely because their visibility is impaired (grass biomass can be 1 g/m2 and exceed 1.5m tall in the woodlands). Thick woody cover in the woodlands conceals stalking predators and tends to cause wildebeest to move less each day than open woodlands. The results suggest that zebra move cautiously through dense woodland patches which can exceed 4km2 and is too far to move in single day, but they move faster through open woodlands. The quality of grass does not affect zebra movement as much as wildebeest; zebra tend to move further each day when the grass is green or in areas with high concentrations of grass nitrogen probably in anticipation of the return movement to the southern plains for the wet season which concurs with the results of other investigators (Brooks and Harris 28, Groom and Harris 21). Wildebeest and zebra depend on access to drinking water, however watering sites are also predictable locations where predators encounter prey (Hopcraft et al. 25, Valeix et al. 29a). Both wildebeest and zebra tend to move directionally when they are near water, which suggests they visit watering sites briefly and do not linger in the vicinity. The daily step lengths are further near water in the woodlands than on the plains, suggesting woodland water sites might be especially dangerous. In conclusion, wildebeest movement is based primarily on optimizing access to sufficient quantities of high quality food, with little regard for predation. Wildebeest herds in excess of 1, animals might dilute the risk of predation for individuals to the point they react less decisively risky habitats. Alternatively, wildebeest might forage in less predatorsensitive manner because they are nutritionally constrained. Zebra movement is based primarily on avoiding predation while accessing the most food of sufficient quality. Both species move further when they are in proximity to humans regardless of other factors, which suggests that migratory herbivores perceive human development as a perpetual threat. The results suggest sympatric species migrate for different reasons and highlights the need for maintaining habitat heterogeneity and continuity for the conservation of migratory species. 154

14 Food and predation differentially affect wildebeest and zebra migrations. LITERATURE CITED Alerstam, T. 26. Conflicting evidence about longdistance animal navigation. Science 313: Anderson, T. M., M. E. Ritchie, E. Mayemba, S. Eby, J. B. Grace, and S. J. McNaughton. 27a. Forage nutritive quality in the Serengeti ecosystem: the roles of fire and herbivory. American Naturalist 17: Anderson, T. M., M. E. Ritchie, and S. J. McNaughton. 27b. Rainfall and soils modify plant community response to grazing in Serengeti National Park. Ecology 88: Balme, G., L. Hunter, and R. Slotow. 27. Feeding habitat selection by hunting leopards Panthera pardus in a woodland savanna: prey catchability versus abundance. Animal Behaviour 74: Bell, R. H. V The use of herb layer by grazing ungulates in the Serengeti. Pages in A. Watson, editor. Animal Populations in Relation to Their Food Resources. Blackwell, Oxford. BenShahar, R., and M. J. Coe The Relationships between Soil Factors, Grass Nutrients and the Foraging Behavior of Wildebeest and Zebra. Oecologia 9: Bivand, R. S., E. J. Pebesma, and V. GómezRubio. 28. Applied spatial data analysis with R. Springer. Bolger, D. T., W. D. Newmark, T. A. Morrison, and D. F. Doak. 28. The need for integrative approaches to understand and conserve migratory ungulates. Ecology Letters 11:6377. Breman, H., and C. T. De Wit Rangeland productivity and exploitation in the Sahel. Science 221:1341. Cromsigt, J., and H. Olff. 26. Resource partitioning among savanna grazers mediated by local heterogeneity: An experimental approach. Ecology 87: Dublin, H. T., A. R. E. Sinclair, and J. McGlade Elephants and fire as causes of multiple stable states in the SerengetiMara Tanzania woodlands. Journal of Animal Ecology 59: Foose, T. J Trophic strategies of ruminant versus nonruminant ungulates. University of Chicago, Chicago. Frair, J. L., E. H. Merrill, H. L. Beyer, and J. M. Morales. 28. Thresholds in landscape connectivity and mortality risks in response to growing road networks. Journal of Applied Ecology 45: Frair, J. L., E. H. Merrill, D. R. Visscher, D. Fortin, H. L. Beyer, and J. M. Morales. 25. Scales of movement by elk (Cervus elaphus) in response to heterogeneity in forage resources and predation risk. Landscape Ecology 2: Fryxell, J. M., J. Greever, and A. R. E. Sinclair Why are migratory ungulates so abundant? American Naturalist 131: Fryxell, J. M., M. Hazell, L. Borger, B. D. Dalziel, D. T. Haydon, J. M. Morales, T. McIntosh, and R. C. Rosatte. 28. Multiple movement modes by large herbivores at multiple spatiotemporal scales. Proceedings of the National Academy of Sciences of the United States of America 15: Fryxell, J. M., J. F. Wilmshurst, and A. R. E. Sinclair. 24. Predictive models of movement by Serengeti grazers. Ecology 85: Fryxell, J. M., J. F. Wilmshurst, A. R. E. Sinclair, D. T. Haydon, R. D. Holt, and P. A. Abrams. 25. Landscape scale, heterogeneity, and the viability of Serengeti grazers. Ecology Letters 8: Chapter 7 155

15 Section II: Balancing Resources and Risk. Grace, J. B., T. M. Anderson, H. Olff, and S. Scheiner. in press. On the specification of structural equation models for ecological systems. Ecological Monographs. Grange, S., P. Duncan, J.M. Gaillard, A. R. E. Sinclair, P. J. P. Gogan, C. Packer, H. Hofer, and M. East. 24. What limits the Serengeti zebra population? Oecologia 14: Harris, G., S. Thirgood, J. G. C. Hopcraft, J. P. G. M. Cromsigt, and J. Berger. 29. Global decline in aggregated migrations of large terrestrial mammals. Endangered Species Research 7: Hebblewhite, M., and E. H. Merrill. 27. Multiscale wolf predation risk for elk: does migration reduce risk? Oecologia 152: Hebblewhite, M., and E. H. Merrill. 29. Tradeoffs between predation risk and forage differ between migrant strategies in a migratory ungulate. Ecology 9: Hebblewhite, M., E. H. Merrill, and T. L. McDonald. 25. Spatial decomposition of predation risk using resource selection functions: an example in a wolfelk predatorprey system. Oikos 111: Hengl, T., G. B. M. Heuvelink, and D. G. Rossiter. 27. About regressionkriging: From equations to case studies. Computers & Geosciences 33: Holdo, R. M., R. D. Holt, and J. M. Fryxell. 29. Grazers, browsers, and fire influence the extent and spatial pattern of tree cover in the Serengeti. Ecological Applications 19:9519. Hopcraft, J. G. C., A. R. E. Sinclair, R. M. Holdo, E. Mwangomo, S. A. R. Mduma, S. Thirgood, M. Borner, J. M. Fryxell, and H. Olff. submitted. Why are wildebeest the most abundant herbivore in the Serengeti? in A. R. E. Sinclair, K. Metzger, S. A. R. Mduma, and J. M. Fryxell, editors. Serengeti IV. Chicago Press, Chicago. Hopcraft, J. G. C., A. R. E. Sinclair, and C. Packer. 25. Planning for success: Serengeti lions seek prey accessibility rather than abundance. Journal of Animal Ecology 74: Kauffman, M. J., N. Varley, D. W. Smith, D. R. Stahler, D. R. MacNulty, and M. S. Boyce. 27. Landscape heterogeneity shapes predation in a newly restored predatorprey system. Ecology Letters 1:697. Kreulen, D Wildebeest habitat selection on the Serengeti plains Tanzania in relation to calcium and lactation a preliminary report. East African Wildlife Journal 13: McNaughton, S. J Ecology of a grazing ecosystem: the Serengeti. Ecological Monographs 55: McNaughton, S. J., J. L. Tarrants, M. M. McNaughton, and R. H. Davis Silica as a defense against herbivory and a growth promotor in African grasses. Ecology 66: Mduma, S. A. R., A. R. E. Sinclair, and R. Hilborn Food regulates the Serengeti wildebeest: A 4year record. Journal of Animal Ecology 68: Morales, J. M., D. Fortin, J. L. Frair, and E. H. Merrill. 25. Adaptive models for large herbivore movements in heterogeneous landscapes. Landscape Ecology 2: Morales, J. M., D. T. Haydon, J. Frair, K. E. Holsiner, and J. M. Fryxell. 24. Extracting more out of relocation data: Building movement models as mixtures of random walks. Ecology 85: Olff, H., M. E. Ritchie, and H. H. T. Prins. 22. Global environmental controls of diversity in large herbivores. Nature 415:9194. Packer, C., R. Hilborn, A. Mosser, B. Kissui, M. Borner, G. Hopcraft, J. F. Wilmshurst, S. Mduma, and A. R. E. Sinclair. 25. Ecological Change, Group Territoriality, and Population Dynamics in Serengeti Lions. Science 37: Pennycuick, L Movements of the Migratory Wildebeest Population in the Serengeti Area between 196 and East African Wildlife Journal 13:

16 Food and predation differentially affect wildebeest and zebra migrations. Reed, D., T. M. Anderson, J. Dempewolf, K. Metzger, and S. Serneels. 28. The spatial distribution of vegetation types in the Serengeti ecosystem: the influence of rainfall and topographic relief on vegetation patch characteristics. Journal of Biogeography. Schick, R. S., S. R. Loarie, F. Colchero, B. D. Best, A. Boustany, D. A. Conde, P. N. Halpin, L. N. Joppa, C. M. McClellan, and J. S. Clark. 28. Understanding movement data and movement processes: current and emerging directions. Ecology Letters 11: Sinclair, A. R. E Does Interspecific Competition or Predation Shape the African Ungulate Community. Journal of Animal Ecology 54: Sinclair, A. R. E., and P. Arcese, editors Serengeti II: Dynamics, management and conservation of an ecosystem. University of Chicago Press, Chicago. Sinclair, A. R. E., and M. Norton Griffiths Does Competition or Facilitation Regulate Migrant Ungulate Populations in the Serengeti Africa. A Test of Hypotheses. Oecologia (Berlin) 53: Sinclair, A. R. E., C. Packer, S. A. R. Mduma, and J. M. Fryxell, editors. 28. Serengeti III: Human Impacts on Ecosystem Dynamics. University of Chicago Press, Chicago. Skoog, R. O Population Ecology of the Plains Zebra in the Serengeti Mara Ecosystem. Texas A & M University and Serengeti Research Institute. Valeix, M., H. Fritz, A. J. Loveridge, Z. Davidson, J. E. Hunt, F. Murindagomo, and D. W. Macdonald. 29a. Does the risk of encountering lions influence African herbivore behaviour at waterholes? Behavioral Ecology and Sociobiology 63: Valeix, M., A. J. Loveridge, S. ChamailleJammes, Z. Davidson, F. Murindagomo, H. Fritz, and D. W. Macdonald. 29b. Behavioral adjustments of African herbivores to predation risk by lions: Spatiotemporal variations influence habitat use. Ecology 9:233. van Soest, P. J Allometry and ecology of feeding behavior and digestive capacity in herbivores: A review. Zoo Biology 15: Wilmshurst, J. F., J. M. Fryxell, B. P. Farm, A. R. E. Sinclair, and C. P. Henschel Spatial distribution of Serengeti wildebeest in relation to resources. Canadian Journal of ZoologyRevue Canadienne De Zoologie 77: Zollner, P. A., and S. L. Lima Search strategies for landscapelevel interpatch movements. Ecology 8: Chapter 7 157

17

18 Food and predation differentially affect wildebeest and zebra migrations. APPENDIX A: WINBUGS CODE model{ for (t in 1:nobs) { # likelihood for steps step[t] ~ dweib(b[id[t]], a[t]) # likelihood for turns ones[t] < 1 ones[t] ~ dbern( wc[t] ) wc[t] < (1/(2*Pi)*(1pow(rho[t],2))/(1+pow(rho[t],2)2*rho[t]*cos(turn[t] mu[t])))/1 turn[t] ~ dunif( , ) logit(rho[t]) < br[id[t]] + br1[id[t]]*dndvi[t] + br2[id[t]]*ndvi[t] + br3[id[t]]*humn[t] + br4[id[t]]*nitgn[t] + br6[id[t]]*wat[t] + br7[id[t]]*biom[t] + br8[id[t]]*vegdist[t] logit(mut[t]) < bm[id[t]] + bm1[id[t]]*dndvi[t] + bm2[id[t]]*ndvi[t] + bm3[id[t]]*humn[t] + bm4[id[t]]*nitgn[t] + bm6[id[t]]*wat[t] + bm7[id[t]]*biom[t] + bm8[id[t]]*vegdist[t] mu[t] < mut[t]* Pi a[t] < exp( ba[id[t]] + ba1[id[t]]*dndvi[t] + ba2[id[t]]*ndvi[t] + ba3[id[t]]*humn[t] + ba4[id[t]]*nitgn[t] + ba6[id[t]]*wat[t] + ba7[id[t]]*biom[t] + ba8[id[t]]*vegdist[t]) } # priors on movement parameters mb ~ dnorm(,.1)i(,) mba ~ dnorm(,.1) mba1 ~ dnorm(,.1) mba2 ~ dnorm(,.1) mba3 ~ dnorm(,.1) mba4 ~ dnorm(,.1) mba6 ~ dnorm(,.1) mba7 ~ dnorm(,.1) mba8 ~ dnorm(,.1) mbr ~ dnorm(,.1) mbr1 ~ dnorm(,.1) mbr2 ~ dnorm(,.1) mbr3 ~ dnorm(,.1) mbr4 ~ dnorm(,.1) mbr6 ~ dnorm(,.1) mbr7 ~ dnorm(,.1) mbr8 ~ dnorm(,.1) mbm ~ dnorm(,.1) mbm1 ~ dnorm(,.1) mbm2 ~ dnorm(,.1) mbm3 ~ dnorm(,.1) mbm4 ~ dnorm(,.1) mbm6 ~ dnorm(,.1) mbm7 ~ dnorm(,.1) mbm8 ~ dnorm(,.1) Chapter 7 159

19 Section II: Balancing Resources and Risk. 16 sb ~ dunif(,1) sba ~ dunif(,1) sba1 ~ dunif(,1) sba2 ~ dunif(,1) sba3 ~ dunif(,1) sba4 ~ dunif(,1) sba6 ~ dunif(,1) sba7 ~ dunif(,1) sba8 ~ dunif(,1) sbr ~ dunif(,1) sbr1 ~ dunif(,1) sbr2 ~ dunif(,1) sbr3 ~ dunif(,1) sbr4 ~ dunif(,1) sbr6 ~ dunif(,1) sbr7 ~ dunif(,1) sbr8 ~ dunif(,1) sbm ~ dunif(,1) sbm1 ~ dunif(,1) sbm2 ~ dunif(,1) sbm3 ~ dunif(,1) sbm4 ~ dunif(,1) sbm6 ~ dunif(,1) sbm7 ~ dunif(,1) sbm8 ~ dunif(,1) tb < 1/(sb*sb) tba < 1/(sba*sba) tba1 < 1/(sba1*sba1) tba2 < 1/(sba2*sba2) tba3 < 1/(sba3*sba3) tba4 < 1/(sba4*sba4) tba6 < 1/(sba6*sba6) tba7 < 1/(sba7*sba7) tba8 < 1/(sba8*sba8) tbr < 1/(sbr*sbr) tbr1 < 1/(sbr1*sbr1) tbr2 < 1/(sbr2*sbr2) tbr3 < 1/(sbr3*sbr3) tbr4 < 1/(sbr4*sbr4) tbr6 < 1/(sbr6*sbr6) tbr7 < 1/(sbr7*sbr7) tbr8 < 1/(sbr8*sbr8) tbm < 1/(sbm*sbm) tbm1 < 1/(sbm1*sbm1) tbm2 < 1/(sbm2*sbm2) tbm3 < 1/(sbm3*sbm3) tbm4 < 1/(sbm4*sbm4) tbm6 < 1/(sbm6*sbm6) tbm7 < 1/(sbm7*sbm7) tbm8 < 1/(sbm8*sbm8) for(i in 1:nind){

20 Food and predation differentially affect wildebeest and zebra migrations. b[i] ~ dnorm(mb,tb)i(,) ba[i] ~ dnorm(mba,tba) ba1[i] ~ dnorm(mba1,tba1) ba2[i] ~ dnorm(mba2,tba2) ba3[i] ~ dnorm(mba3,tba3) ba4[i] ~ dnorm(mba4,tba4) ba6[i] ~ dnorm(mba6,tba6) ba7[i] ~ dnorm(mba7,tba7) ba8[i] ~ dnorm(mba8,tba8) br[i] ~ dnorm(mbr,tbr) br1[i] ~ dnorm(mbr1,tbr1) br2[i] ~ dnorm(mbr2,tbr2) br3[i] ~ dnorm(mbr3,tbr3) br4[i] ~ dnorm(mbr4,tbr4) br6[i] ~ dnorm(mbr6,tbr6) br7[i] ~ dnorm(mbr7,tbr7) br8[i] ~ dnorm(mbr8,tbr8) bm[i] ~ dnorm(mbm,tbm) bm1[i] ~ dnorm(mbm1,tbm1) bm2[i] ~ dnorm(mbm2,tbm2) bm3[i] ~ dnorm(mbm3,tbm3) bm4[i] ~ dnorm(mbm4,tbm4) bm6[i] ~ dnorm(mbm6,tbm6) bm7[i] ~ dnorm(mbm7,tbm7) bm8[i] ~ dnorm(mbm8,tbm8) } Pi < } Chapter 7 161

21 Section II: Balancing Resources and Risk. APPENDIX B. ANIMAL CAPTURE, HANDLING AND GPS RADIO COLLARS Female animals older than 2 and younger than 5 (based on tooth ware) were randomly selected from mixed herds dispersed over the plains of Serengeti. All animals were immobilized by Tanzania National Parks or Tanzania Wildlife Research veterinarians using approximately 7 to 8mg of Etorphine (M99) and reversed using Diprenorphine (M55). The total time the animals were immobilized for did not extend beyond 15 minutes and all females were quickly reunited with the herds or harems (in the case of zebra). While immobilizing females with offspring, care was taken not to separate mothers for extended periods. Offspring would generally wait near by while their mothers were immobilized, and all females almost immediately found their calves (or foals) upon recovery. The average neck circumference for female wildebeest was 66cm, and 68cm for female zebra. Lotek GPS1 collars with VHF tracking and a UHF modem for remotely downloading data were used on wildebeest from 1999 to 24. From 25 to 26 we used Televilt SatLink collars with GPS data transmission via satellite phone for wildebeest and zebra. Televilt Tellus Simplex with a GSM modem for data transmission were used from 26 to 28 for both species, and proved to be the most reliable. GPS locations of the animals collected at :, 6:, 12: and 18: daily from 1999 to 24 and at 6: and 18: daily from 25 onwards. 162

22 Food and predation differentially affect wildebeest and zebra migrations. APPENDIX C: ADDITIONAL INFORMATION FOR MATERIAL AND METHODS Food quality The quality of the forage was estimated from grass nitrogen measured at 148 sites (Figure 1) distributed in the different soil and vegetation types in Serengeti and across the rainfall gradient. The aboveground grass biomass was clipped in five 25cm x 25cm plots at each of the 148 sites (pooling all grass species) and air dried before being ground in a cyclonic grinder (Foss Cyclotec 193) with a sieve size of 2mm. Ground samples were further oven dried for 48 hours at 72 C and nitrogen concentrations were measured using a Near InfraRed (NIR) spectrophotometer (Bruker MPA NIR). The NIR estimation of grass nitrogen concentrations proved to be a fast and accurate technique. The correlation between the true nitrogen concentration as established from a CHNS elemental analyzer (CarloErba Instruments EA111 CHNS) and the predicted NIR nitrogen concentrations of a subset of samples suggest the NIR technique is highly accurate (r2 =.97, n = 76). The spatial distribution of grass nitrogen for the ecosystem was interpolated by regression kriging. Regression kriging (Hengl et al. 27, Bivand et al. 28) is an interpolation technique that takes into account the spatial autocorrelation between sampling points (as per ordinary kriging), while simultaneously accounting for the correlation between samples and an underlying predictor variable. Since grass nitrogen is inversely correlated with NDVI (r2 =.1, slope =.1, p<.1, n = 148), we used the 9year mean NDVI index (229) as a covariate to more accurately predict the spatial distribution of grass nitrogen. As a separate internal accuracy check, 3 samples were selected randomly and used to test the accuracy of a second regression kriged grass nitrogen map created from the remaining 118 samples. The predicted grass nitrogen values from the second map correlated well with the 3 random samples (r2 =.25, slope =.48, p <.1, n = 3), and verify this technique as having an acceptable level of accuracy. Normalized Difference Vegetation Index (NDVI) is commonly used as a measure of the vegetation greenness; NDVI measures approaching 1 indicate live green vegetation while those approaching 1 suggest dry or dead vegetation. Sixteenday NDVI composites at resolution of 25m were downloaded from the MODIS Terra satellite for the period from 2 to 29 using NASA s Warehouse Inventory Search Tool website (WIST, NDVI was used as a predictor of the movement patterns of wildebeest and zebra by (a) estimating the amount of green vegetation available to grazers, and (b) to estimate the rate at which the grasses are greening or drying. Grasses tend to flush and senesce rapidly with rainfall and evapotranspiration, therefore large differences between current and previous NDVI scenes suggest patches are either rapidly drying or greening (positive or negative values, respectively). Conversely, small differences between the current and previous 16day NDVI composite suggest the area is stable and neither drying nor greening. Chapter 7 Food abundance Grass biomass is positively correlated with rainfall and soil moisture. Under mesic conditions grasses grow tall and tend to become lignified which decreases their digestibility and their nutritional value to herbivores (Breman and De Wit 1983, McNaughton et al. 1985, Olff et al. 22). Additionally, the species composition of grasses shifts across the mesicarid gradient such that mesic areas tend to be dominated by lower quality species (Anderson et al. 27). Therefore, we used rainfall and the topographic wetness index (TWI) as a proxy for grass biomass and not grass quality. The longterm average rainfall was interpolated by regression kriging monthly rainfall records from 58 gauges for 46 years across a known southeast to northwest diagonal rainfall 163