1. Introduction

For most mammalian predators prey size increases with body size, with small carnivores (<20 kg) predating on prey smaller than themselves and large carnivores (>20 kg) predating on prey their own size or larger (Carbone et al., 1999, 2007). Although a shift from small to large prey increases energy intake, pursuing and subduing large prey requires more energy and therefore results in an increase in energy expenditure. Consequently, carnivores have developed several behavioural and ecological strategies to optimise energy intake and expenditure (Carbone et al., 2007). African wild dogs (Lycaon pictus) are social coursing predators with a body mass of ca. 25 kg (Carbone et al., 2007 (S1)). By hunting cooperatively in packs the species is able to reduce foraging costs and pursue and subdue prey much larger than their own body size (Creel & Creel, 2002; Radloff & du Toit, 2004). African wild dogs generally predate on antelopes within a body mass range of 16–32 kg and 120–140 kg (Hayward et al., 2006a). Preferred prey species in east and southern Africa are greater kudu (Tragelaphus strepsiceros), Thomson’s gazelle (Gazella thomsonii), impala (Aepyceros melampus) and bushbuck (Tragelaphus scriptus) (Hayward et al., 2006a).

African wild dogs hunt daily, at dusk and dawn or when sufficient moonlight is available (Estes & Goddard, 1967; Creel & Creel, 2002). The species hunts opportunistically using sight, aided by auditory and olfactory cues (Estes & Goddard, 1967; Creel & Creel, 2002). Hunting strategy varies with habitat type, with long-distance pursuits of prey by multiple individuals in a pack in open grassland habitats (Estes & Goddard, 1967), and several individual hunting attempts, which result in low individual kill rates but high pack feeding rates as prey is shared, in mixed woodland habitat (Hubel et al., 2016). As cursorial hunters, African wild dogs show a tendency to predate on the slowest individuals, e.g. young animals or animals in poor body condition (Fitzgibbon & Fanshawe, 1989; Pole et al., 2004).

While hunting, carnivores have more at stake than energetic expenditure and hunting success alone: they can get injured or killed by their prey (Mukherjee & Heithaus, 2013; Kerley, 2018). Therefore, deciding when to attack or avoid potentially dangerous prey is an important aspect of prey selection (Mukherjee & Heithaus, 2013). Like other carnivore species (Hayward et al., 2006b), African wild dogs seem to take the probability of injury into account by selecting prey species which pose minimal risk of injury and avoiding prey species which are likely to stand and defend themselves (Hayward et al., 2006a).

Baboon (Papio spp.) is a prey species which can inflict serious injury and mortality (Cowlishaw, 1994; Estes, 2004) and, where possible, Africa’s large carnivores usually avoid predation on baboons (Hayward & Kerly, 2005; Hayward, 2006; Hayward et al., 2006a, b). Baboons have occasionally been recorded as part of the African wild dog’s diet (Woodroffe et al., 2007; Mbizah et al., 2012), however, as far as we are aware, observations of hunts on this prey species have never been described (see also Cowlishaw, 1994; Hayward et al., 2006a). Although baboon lies within their preferred prey body mass range, African wild dogs seem to actively avoid predation on this dangerous prey species (Hayward et al., 2006a). In this study we investigate prey selection patterns of African wild dogs in Mana Pools National Park, Zimbabwe, and discuss which ecological factors are likely to contribute to baboon hunting by African wild dogs.

2. Method

2.1. Study area

Mana Pools National Park (MPNP) is a 2196 km² UNESCO Natural World Heritage Site in northern Zimbabwe (15°56′S, 29°27′E), managed by the Zimbabwe Parks and Wildlife Management Authority. Mean annual rainfall is ca. 706 mm with a wet season from November to March. July is the coldest month (mean minimum temperature 10.7°C), October the hottest (mean maximum temperature 39.7°C). Four main vegetation types are found in MPNP: Colophospermum mopane woodland, Faidherbia albida woodland, Brachystegia–Julbernardia woodland and Commiphora–Combretum thicket (Figure 1). MPNP is utilised for photographic safaris by vehicle, on foot or by canoe (ZPWMA, 2009).

View Full Size

Map of Mana Pools National Park, showing the area along the floodplain of the Zambezi river where data on the two African wild dog study packs, their kills and prey availability were collected.

Citation: Behaviour 156, 1 (2019) ; 10.1163/1568539X-00003529

• Download Figure
• Download figure as PowerPoint slide

MPNP is situated along the Zambezi river (Figure 1), which is the main permanent water source for the park. The ecology of MNNP’s large mammals is affected by a seasonal dry-season concentration of wildlife on the valley floor due to water availability and the presence of food sources due to prolonged plant productivity and availability of Faidherbia albida on the alluvial floodplains (ZPWMA, 2009). MPNP is home to all species within Africa’s large carnivore guild: lion (Panthera leo), spotted hyena (Crocuta crocuta), leopard (Panthera pardus), cheetah (Acinonyx jubatus), brown hyena (Parahyaena brunnea) and African wild dog (van der Meer, 2016), and harbours a variety of herbivore species, such as buffalo (Syncerus caffer), eland (Tragelaphus oryx), impala, greater kudu, waterbuck (Kobus ellipsiprymnus) and zebra (Equus quagga). The area is also home to a population of chacma baboons (Papio ursinus) (WEZ, 2017).

2.4. Analyses

We used a Pearson’s chi-square test to determine whether there was an association between the number of impala and baboon kills recorded for each pack and the data collection method: direct observations, observations from BBC footage and analyses of faecal samples. As we found no significant association (pack I: ${\mathit{\chi }}^2=1.12$, $\mathit{p}=0.57$, pack II: ${\mathit{\chi }}^2=0.37$, $\mathit{p}=0.83$) prey selection patterns were analysed by combining the three data sets. Duplicate sightings were removed by considering sightings of the same pack, on the same day, within the same time frame, killing the same prey species to be one sighting. Passage rate of food items through the gastrointestinal tract has been predicted to be 24 hours (Davies-Mostert et al., 2010). Faecal samples collected 24 hours after an observation of a kill which contained hair of a species similar to the prey species killed were considered duplicates and removed from the analyses. A total of 153 kills were included in the final analyses (direct observations ($\mathit{N}=29$), BBC footage ($\mathit{N}=94$), faecal samples ($\mathit{N}=30$)).

Data on availability of prey species in the area covered by the two African wild dog packs (Figure 1) were obtained from Wildlife and Environment Zimbabwe’s annual game counts (WEZ, 2015, 2016, 2017). During these dry season counts experienced game count teams (including a Zimbabwe Parks and Wildlife game ranger) walk across the ca. 45 km² central Mana Pools floodplain (Figure 1) on north-south transect lines which are 500 m apart and record bird, carnivore and herbivore species within the 250 m zone on either side of the transect. In two consecutive days, each transect is walked four times: twice in the early morning and twice in the late afternoon (WEZ, 2017). For each year we calculated the proportion of available African wild dog prey species along the floodplain by dividing the mean count of that species (total number counted of a prey species divided by the number of transects) by the sum of the mean count for all prey species (Table 1).

View Full Size

Annual mean number and standard error of African wild dog prey species counted during the transect walks at the annual WEZ game count and the proportion of these species in the prey population along the MPNP floodplain.

Citation: Behaviour 156, 1 (2019) ; 10.1163/1568539X-00003529

• Download Figure
• Download figure as PowerPoint slide

Following Hayward et al. (2006a), we used a Jacobs’ index (Jacobs, 1974) to determine whether African wild dogs predated on prey species relative to availability or whether they avoided or preferred certain prey species. The Jacobs’ index follows the equation $\mathit{D}=\mathrm{(}\mathit{r}-\mathit{p}\mathrm{)}\mathrm{/}\mathrm{(}\mathit{r}+\mathit{p}-2\mathit{r}\mathit{p}\mathrm{)}$, where r is the proportion of prey predated on and p the proportion of prey available. D ranges between −1 (strong avoidance) and +1 (strong preference). In accordance with Hayward et al. (2011), we considered Jacobs’ index values between −0.2 and 0.2 as an indication the prey species was predated on as expected based on its availability. Jacobs’ index was calculated per pack per year, using the available proportion of prey for a given year. In addition, we calculated the percentages of prey taken within each age and sex class and used a binomial test (test proportion 0.5) to determine whether certain age and sex classes were more often predated on.

Using a genlin procedure, we fitted binary logistic regression models to determine which factors affect the number of baboon kills. Where appropriate we first used a non-parametric, two tailed, point-biserial correlation to test for correlations between variables. Treating impala kills as the reference, we modelled the number of baboon kills in relation to season (wet or dry), breeding season (denning: when pups are too small to follow the pack and the pack returns to the den after foraging, or nomadic: when the pups follow the pack) and time of day the kill was made (unknown, AM or PM). Because pack identity and pack size (number of adults and subadults) were significantly correlated ($\mathit{r}=-0.43$, $\mathit{p}\mathrm{<}0.01$), only pack identity was included in the models. We considered all possible models without interactions and used Akaike’s Information Criterion adjusted for small sample size (AICc) to rank models. Models within 2 AICc units of the best model were considered to have strong support (Burnham & Anderson, 2002). For each parameter included in the confidence set of models, we computed cumulative AICc weights (${\mathit{\omega }}_{\mathit{i}}$) across the models which contained the variable of interest and calculated a model averaged parameter estimate and 85% confidence intervals (Arnold, 2010) to evaluate its importance (Burnham & Anderson, 2002; Arnold, 2010). Statistical analyses were performed with SPSS software version 20.0 (SPSS, Chicago, IL, USA).

3. Results

We recorded a total of 113 kills by pack I (2015 $\mathit{N}=32$, 2016 $\mathit{N}=49$, 2017 $\mathit{N}=32$) and 40 kills by pack II (2015 $\mathit{N}=13$, 2016 $\mathit{N}=14$, 2017 $\mathit{N}=13$). Both study packs predominantly preyed upon impalas and baboons. Pack I was recorded to kill impalas ($\mathit{N}=67$), baboons ($\mathit{N}=45$) and a buffalo calf ($\mathit{N}=1$), pack II to kill impalas ($\mathit{N}=29$), baboons ($\mathit{N}=8$) and subadult warthogs ($\mathit{N}=3$). Throughout the study period, the annual availability of baboons and impalas in the study area remained relatively stable (Table 1). Based on the Jacobs’ index calculated per study year, pack I less frequently killed impalas and more frequently killed baboons than expected based on availability (Figure 2a). Pack II predated on impalas relative to availability. Although the pack initially took fewer baboons than expected based on availability (2015), in later years (2016–2017) baboons became a preferred part of the diet (Figure 2b).

View Full Size

Jacobs’ index for prey species predated on by two study packs of African wild dogs in Mana Pools National Park, +1 indicates strong prey preference, −1 indicates strong prey avoidance while values between −0.2 and 0.2 indicate prey is predated on relative to its availability.

Citation: Behaviour 156, 1 (2019) ; 10.1163/1568539X-00003529

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Pack of African wild dogs feeding off adult male baboon kill (a) and African wild dog catching a subadult female baboon (b) Photo credit: BBC.

Citation: Behaviour 156, 1 (2019) ; 10.1163/1568539X-00003529

• Download Figure
• Download figure as PowerPoint slide

The study packs more often predated on adult than subadult impalas ($\mathit{N}=60$, $\mathit{p}\mathrm{<}0.01$) and took more female than male impalas ($\mathit{N}=56$, $\mathit{p}\mathrm{<}0.01$). Where impala age and sex could be determined ($\mathit{N}=60$), adult females (61.7%) outnumbered adult males (21.7%), followed by juveniles (6.7%) and subadult males (5.0%) and females (5.0%). The study packs predated as often on adult as on subadult baboons ($\mathit{N}=34$, $\mathit{p}=1.00$), and did not take significantly more male than female baboons ($\mathit{N}=10$, $\mathit{p}=0.75$). Where baboon age and sex could be determined ($\mathit{N}=9$), five were adult males (Figure 3a), three were adult females and one was a subadult female (Figure 3b). Nine adults, 16 subadults and two juveniles were of unknown sex.

View Full Size

Variables, number of parameters in the model (K), Akaike’s Information Criterion adjusted for small sample size (AICc), differences in AICc value between the model and the model with the lowest AICc value (ΔAICc), AICc weight (${\mathit{\omega }}_{\mathit{i}}$) and log likelihood (LL) for models used to predict which factors determine the number of baboon kills.

Citation: Behaviour 156, 1 (2019) ; 10.1163/1568539X-00003529

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Model averaged parameter estimates (β), standard errors (SE), 85% confidence intervals (CI) (see Arnold, 2010) and Akaike importance weights ($\sum {\mathit{\omega }}_{\mathit{i}}$) for the parameters in the highest ranking models (ΔAICc ⩽ 2) used to predict which factors determine the number of baboon kills.

Citation: Behaviour 156, 1 (2019) ; 10.1163/1568539X-00003529

• Download Figure
• Download figure as PowerPoint slide

The parameter time of day the kill was made did not occur in the confidence set of models and was therefore considered not to have a substantial impact on the number of baboon kills. Based on the models with the most support (Table 2), the relative importance of individual parameters (Table 3) and taking into account 85% confidence intervals of the model averaged parameter estimates, the number of baboon kills was higher in the dry season than the wet season and lower in the denning season than the nomadic season. Model averaged 85% confidence intervals for the parameter pack identity included zero (Table 3), this parameter was therefore discarded as an uninformative parameter (Arnold, 2010). During the study period, pack size (adults and subadults) varied considerably (Table A1 in the Appendix) with an overall average pack size of 16.0 ± 1.5 (mean ± SE) adults and subadults for pack I and 10.9 ± 0.4 (mean ± SE) adults and subadults for pack II. The point-biserial correlation confirmed there was no significant relationship between pack size and the number of impalas or baboons killed ($\mathit{r}=-0.04$, $\mathit{p}=0.60$).

Direct observations ($\mathit{N}=29$) showed spotted hyenas were present at 20.7% of the kills (four impala kills, one baboon kill and a kill of a buffalo calf). However, losses to kleptoparasitism were little as spotted hyenas generally only arrived after the African wild dogs were close to finishing or had finished feeding from their kill.

4. Discussion

Although the line transect method is a flexible and cheap method to estimate the abundance of wildlife species, it has its shortcomings (Greene, 2012). Especially when, like in this study, limitations of the data set prevent the use of distance sampling software to account for variation in detectability. Besides, because this method assumes all animals are detected at their initial location, animal movement can create bias, particularly when animals move in response to the observers (Palka & Hammond, 2001; Glennie et al., 2015). As a result of allowing unguided walks, the wildlife in MPNP is exceptionally relaxed about human presence. In addition, the floodplain consists of very open vegetation which facilitates easy detection of wildlife. Counting bias due to differences in detection ability or animal movement is therefore minimal. Game counts took place in the dry season in an area containing the main permanent water source of MPNP. Although less so for browers, distribution patterns of grazers are strongly affected by water availability (Redfern et al., 2003). The game count data set is therefore unlikely to be representative for prey abundance in MPNP as a whole. However, because the area sampled is representative for the area utilised by the African wild dog study packs (Figure 1) we do feel the game count data provides a reliable indication of relative prey availability which can be used to draw conclusions about prey selection of the two study packs.

The African wild dog packs in our study predominantly predated on impalas and baboons. Compared to impalas, which present limited threat of injury to predators (Hayward et al., 2006a, b), baboons, a prey species for which adult males especially are known to retaliate against predators, present a high risk of injury (Cowlishaw, 1994; Estes, 2004). It is challenging to accurately assess frequency of injury from dangerous prey (Mukherjee & Heithaus, 2013) and our data on frequency of baboon related injuries are limited. However, we observed several cases in which baboons defended themselves, their young, or other troop members against African wild dog attacks (Figure 4) and baboon hunts did result in injury (Figure 5). Injury is an important foraging cost (Mukherjee & Heithaus, 2013) which predators take into account by avoiding prey which presents a high risk of injury or mortality (Hayward et al., 2006a, b). However, in this study, once included in the diet, the African wild dog study packs preferred to predate on baboons.

View Full Size

Stand-off between female baboon and African wild dog which attempts to predate on a juvenile baboon (a) and adult male baboon chasing African wild dog (b). Photo credit: BBC.

Citation: Behaviour 156, 1 (2019) ; 10.1163/1568539X-00003529

• Download Figure
• Download figure as PowerPoint slide

View Full Size

African wild dog with baboon injury on left flank (a) and African wild dog with baboon injury on its back (b). Photo credit: BBC.

Citation: Behaviour 156, 1 (2019) ; 10.1163/1568539X-00003529

• Download Figure
• Download figure as PowerPoint slide

Prey vulnerability to predation varies over age and sex classes (Cowlishaw, 1994; Pole et al., 2004) and is affected by season (Pole et al., 2004). While some studies describe adult male baboons to be most vulnerable to predation (Cowlishaw, 1994), others only found age differences in vulnerability (adults more vulnerable than subadults) (Jooste et al., 2012). The African wild dog packs in this study predated equally on male and female and adult and subadult baboons. Vulnerability of impala sex and age classes of to African wild dog predation is affected by sex ratio and season (Reich, 1981; Pole et al., 2004). In some ecosystems, overall predation is higher on adults males (Pole et al., 2004) while in other ecosystems female and adult impalas are more vulnerable to African wild dog predation (Reich, 1981; Vogel et al., 2018). Compared to males and subadults, our study packs predated more heavily on female and adult impalas. Predation on impalas increased in the wet season, baboon predation was higher in the dry season. Baboons do not have a demarcated breeding season (Cheney et al., 2004) but impalas give birth to their fawns in the wet season (Child, 1964). During this period, pregnancy and lactations reduces fat reserves of female impalas to a minimum (Pole et al., 2004). African wild dogs select young prey animals or prey in poor body condition (Fitzgibbon & Fanshawe, 1989; Pole et al., 2004) and the wet season provides an abundance of out of condition impala females and young and naïve impalas that are relatively easy to catch (Pole et al., 2004). In addition, extensive elephant movement along the muddy floodplain of MPNP results in a rugged surface of hardened, dried out, elephant footprints in the dry season, herewith increasing the difficulty of high speed pursuits of impalas and the indirect risk of injury such pursuits present (Mukherjee & Heithaus, 2013).

Baboon kills were less frequently made in the denning season when the need to feed the pups at the den can affect prey selection (Woodroffe et al., 2007). Body mass of adult impalas (adult male ca. 60 kg, adult female ca. 45 kg) is 2–3 times higher than body mass of adult baboons (adult male ca. 32 kg, adult female ca. 15.4 kg) of the same sex (Estes, 2004), a baboon kill is therefore less likely to feed the entire pack. Although African wild dogs generally consume 2–3 kg of meat per individual (Creel & Creel, 1995; Carbone et al., 2007), stomach capacity is ca. 9 kg (Creel & Creel, 1995). Consequently, predation on impalas maximizes food intake of the pack and herewith the quantity of meat which can be regurgitated for the pups upon return to the den. With larger packs being able to kill larger prey (Creel & Creel, 2002), a preference for smaller prey could be an indication of suboptimal pack sizes. However, in accordance with Woodroffe et al. (2007), we found no evidence pack size (adults + subadults) affected predation on baboons or impalas. Throughout the study, pack size of our African wild dog study packs was ⩾7 adults, which is considered a viable pack size for cooperative hunting and breeding (Courchamp et al., 2002). Therefore, an inability to secure larger prey due to suboptimal pack sizes does not explain why African wild dogs in MPNP preferentially predated on baboons.

Preferential predation on baboons also does not seem to be the result of a low availability of impalas. There are many ecosystems where impalas are the principal prey species of African wild dogs, compared to these systems, impala densities along the MPNP floodplain (37.3/km²) are relatively high (e.g., Selous 28.6/km², Savé 11.7/km²) (Creel & Creel, 2002; Pole et al., 2004). Apart from the presence of prey, prey selection may also be affected by the presence of larger competitors: the risk of encountering larger competitors increases with handling time of the kill, which is shorter for smaller prey items (Hamilton, 2002). However, compared to other ecosystems (presence larger competitors at kill sites 9–86%, kleptoparasitism 0–100%), kleptoparasitism and presence of larger competitors at kill sites of African wild dogs was not exceptionally high (Creel & Creel, 1998; van der Meer et al., 2011); therefore, interspecific competition alone does not seem to explain prey selection by the MPNP study packs.

Profitability of prey is related to energy content but also to energy spend finding, pursuing and subduing prey (Creel & Creel, 2002). Like in other ecosystems, along the MPNP floodplain impalas and baboons have formed interspecific associations to optimise feeding and anti-predator strategies (Davis & Ebersole, 2015). Consequently, search efficiency for impalas and baboons are likely to be similar; however, chase distance may differ substantially. When encountering a predator, impalas are quick to take flight, while baboons, when foraging on the ground during daytime, seek refuge in the nearest tree and when necessary confront a predator (Cowlishaw, 1997; Estes, 2004). The MPNP floodplain consists of open woodland with little understory in which it is relatively easy to detect approaching predators and difficult to find refuge. Such open habitat increases flight initiation distance (Stankowich & Blumstein, 2005) which increases the distance over which prey is chased by African wild dogs (Reich, 1981) and the likelihood of a kill (Creel & Creel, 2002). Where impalas are known for their manoeuvrability and high speed (ca. 50 km/h) when escaping predation (Wilson et al., 2018), baboons are much slower. In this study, baboons were always chased over short distances (<100 m) and, unlike impala hunts, baboon hunts would normally result in the pack outrunning and killing at least one individual in the troop before it found refuge. Such differences in hunting efficiency can outweigh a lower energy intake and enable African wild dogs to subsist on smaller prey species (Woodroffe et al., 2007).

As far as we are aware, this is the first study which describes preferential predation of African wild dogs on baboons (see also Cowlishaw, 1994; Hayward et al., 2006a). Hunting dangerous prey like baboons requires skills which vary among individuals (Jooste et al., 2013). Baboon hunting in MPNP is believed to have started with the original alpha female of pack I, which was first witnessed to hunt baboons about a decade ago (personal communication safari operators). Both study packs consist of offspring of this female. Although pack II initially avoided baboon predation, once seven dispersers of pack I joined the pack in December 2015, baboons became a preferred part of the diet. In addition to these two study packs, there are five known African wild dog packs in MPNP, only one of these other packs has ever been witnessed to predate on baboons (personal communication safari operators), this pack was formed mid 2017 and consists of four individuals of which two are dispersers from pack I. In accordance with Jooste et al. (2013), these observations suggest predation on baboons may be an individually acquired pattern of behaviour which disseminates through the MPNP African wild dog study packs via social learning (acquiring new behaviour by observing and imitating others, see Galef, 1976). However, further research will be necessary to confirm this hypothesis and to gain a more detailed insight in the various factors likely to affect the MPNP study packs’ unusual prey preference for baboons.