As human populations expand, leading to agriculture, settlements, and roads, wild dogs are losing the spaces in which they were once able to roam freely. In the Hamburg landscape, AWF, with funding from the U.S. Agency for International Development and the Dutch government employed 12 scouts from five neighboring communities.
By providing access to new employment, AWF is able to weave conservation and economic opportunity together to incentivize wild dog protection. Wild dogs also have a large range of vocalizations that include a short bark of alarm, a rallying howl, and a bell-like contact call that can be heard over long distances.
The hunting members of the pack return to the den where they regurgitate meat for the nursing female and pups. When pack numbers are reduced, hunting is not as efficient, and adults may not bring back sufficient food for the pups.
They hunt for a wide variety of prey, including gazelles and other antelopes, warthogs, wildebeest calves, rats, and birds. Like most predators, they play an important role in eliminating sick and weak animals, thereby helping maintain the natural balance and improve prey species.
The Africanism Dogs scientific name ‘Lyon ictus’ comes from the Greek language for ‘wolf’ and Latin for ‘painted’. The irregular pattern is colored with white, yellow, brown and black markings.
Africanism Dogs have slim, lean bodies and long, slender legs. They have large, rounded distinctive ears and a long tail which has a white plume at the end.
East and West African dogs tend to be smaller than those in South Africa. Africanism Dogs differ from other members of the candidate family in that they only have four toes on each paw instead of five as they lack dew claws (which is the fifth digit on other can ids).
Africanism Dogs have around 42 teeth including premolars that are much larger than in other can ids allowing it to consume large amounts of bone. They prey upon a variety of grazing animals particularly medium-sized ungulates such as Zebras, Antelopes, Impalas, Gazelles and Springboks.
Wild dogs are very sociable animals and have a submissive based hierarchy rather than a dominant one. This non-aggressive approach is emphasized perhaps because if any injuries occur, the pack will be short of hunters and unable to provide as much for its members.
They are extremely co-operative as a hunting pack when running down and over-powering prey in long distance chases. In the early, cool mornings and late afternoons the Wild Dogs will approach their prey in full view.
Surprise attacks are unnecessary as Africanism Dogs have the stamina to chase prey until it is exhausted. During these long distance chases, Wild Dogs will spread out to prevent prey from any sideways escape attempts.
The preys zigzagging evasive movements which would normally confuse a lone hunter such as a Cheetah, are ineffective against the pack of wild dogs. As the exhausted prey eventually slows down, the dogs surround it targeting their softer underparts and killing their victim.
While a whole herd of ungulates may be targeted, the eventual victim will be the one who falls behind due to age or sickness. Africanism Dogs have a very powerful bite and their large molars and premolars allow them to easily crush the bones of their catch.
After a gestation period of around 70 days, the female gives birth to a litter of around 10 pups (few usually survive because of predators). Pups are weaned at 10 weeks and when they reach 3 months, they leave the den to begin running with the pack.
Competition with larger carnivores such as lions and spotted hyenas is also a problem for the wild dog as they both pursue the same type of prey. Wild dogs are also killed by farmers who want to protect their livestock and disease can spread from domestic animals.
Broad-scale models describing predator prey preferences serve as useful departure points for understanding predator-prey interactions at finer scales. Previous analyses used a subjective approach to identify prey weight preferences of the five large African carnivores, hence their accuracy is questionable.
Based on simulations of known predator prey preference, for prey species sample sizes above 32 the segmented model approach detects up to four known changes in prey weight preference (represented by model break-points) with high rates of detection (75% to 100% of simulations, depending on number of break-points) and accuracy (within 1.3±4.0 to 2.7±4.4 of known break-point). An assessment of carnivore diets throughout Africa found these accessible prey weight ranges include 88±2% (cheetah), 82±3% (leopard), 81±2% (lion), 97±2% (spotted hyena) and 96±2% (wild dog) of kills.
These descriptions of prey weight preferences therefore contribute to our understanding of the diet spectrum of the five large African carnivores. Where datasets meet the minimum sample size requirements, the segmented model approach provides a means of determining, and comparing, the prey weight range preferences of any carnivore species.
This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. CJ Gambling was funded by the National Research Foundation and a Claude Leon Postdoctoral Fellowship.
The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Carnivore numbers are declining globally, with a reduction in distribution and abundance leading to almost a quarter of species now being threatened with extinction .
However, the method employed by these studies to determine preferred prey weight ranges is subjective. The weight range representing the peak of this curve was identified as the “most preferred” weight range by linking positive Jacobs’ Index values to prey body mass values on the x-axis .
However, this approach does not offer a statistically independent means of determining where on the peak of the curve the “most preferred” weight range lies, and is therefore subjective. Furthermore, extracted values may misrepresent actual preferred prey weight ranges if the coefficient of determination of the model fit is low, indicating that outlying values are influencing model fit (e.g. wild dog ; ).
The accuracy of this method is therefore questionable, and it is neither replicable nor comparable across carnivore species. Despite the questionable accuracy of these weight ranges, they are serving as a point of departure for a wide range of applications, including ecological studies , , conservation suggestions and human-wildlife conflict issues .
This approach is shown to allow for a more complete description of each predator’s prey weight spectrum. Commonly used preference indices such as the forage ratio and Inlet’selectivity index suffer from non-linearity, bias towards rare food items, increasing confidence intervals with increasing heterogeneity, being unbound or undefined and lacking symmetry between selected and rejected values .
The kill data collected in these studies were derived from both incidental observations and continuous follows, as well as from scat analyses in the case of leopard and spotted hyena. While incidental observations are biased toward larger prey; this bias against smaller items is generally alleviated by the under counting of small prey species in aerial counts .
The sources, location and timing of African studies used by Hayward and colleagues and the number of kills recorded in each are presented in Table S1. For each carnivore, prey abundance and kill data were used to calculate a J.I.
Value for each listed prey species at each site detailed in Table S1, as done in the original preference papers. Each potential prey species was allocated a standard species mass of three-quarters of the mean adult female body mass , in order to account for calves and sub-adults eaten (, ; Table S2).
This assumption appears robust when tested with kill data for leopard . First we consider a situation in which a predator displays equal preference for all 10 species.
If a predator preferred the species weighing between 51 and 55 kg twice as much as the species weighing more than 55 kg, the slope of the relationship between prey mass and cumulative preference would be twice as steep for the first five prey masses than for the second five (Fig. In this case the relationship between the response and explanatory variables would be piece wise linear (segmented), represented by two straight lines connected at a “break-point” .
In (c) prey masses are represented by categorical values to generate even mass distribution and the predator prefers prey weighing 51 to 55 kg twice as much as those weighing more than 55 kg. In order to control for this, prey species masses can be ranked from lightest to heaviest and converted into categorical integer values of equal increment (from 1 to 10), in order to ensure that all masses have equal weighting.
In order to assess the accuracy of the proposed segmented model approach in determining a predator’s prey weight preferences, and to determine minimum sample size requirements, we simulated predator diets with known changes in prey preference (and thus break-points). For each simulation, we randomly selected a total prey species sample size between six and 100, and these species were randomly allocated between the preference groups (between two and five preference groups).
Therefore, the accuracy of break-point detection can be assessed as a function of the total minimum sample size, determined by multiplying the minimum sample in the smallest preference group by the number of groups. Break-point detection accuracy was assessed by calculating, for each simulation which met the minimum sample size requirements, the absolute difference in prey species categorical value between the known break-point and the detected break-point.
These models were conducted in the open source statistical package R (R Development Core Team 2012) using the segmented generalized linear model function in the “segmented package” ( ; see example code detailed in Code S1). The segmented model approach was performed for each of the five large carnivores (separately), using the data transformations detailed below.
Values were available from two or more sites) were therefore ranked from lightest to heaviest according to standard species masses and each species was allocated an integer value, commencing at 1 for the lowest prey mass (Table S2). The optimum number of break-points (where more than one existed) was therefore selected using Alike’s information criterion (AIC; ).
As a segmented model was the best-fit function for all five carnivores, mass-ranks at which the best-fit model detected break-points were then translated back into actual prey masses (standard species masses; Table S2). Where a break-point fell directly on a mass-rank, the species mass corresponding to that mass-rank was included in the weight range to the left of this break-point.
Within each of the weight ranges identified by model break-points, the actual degree of prey preference was quantified. Value of each prey weight range across sites was tested for significant preference or avoidance using a single sample t-test against a mean of zero where data conformed to the assumptions of normality, and a Wilcox on signed-rank test where data did not .
Value significantly greater than zero indicated a preferred prey weight range, a mean J.I. Value not significantly different from zero indicated prey in a weight range killed relative to their abundance and a mean J.I.
Value significantly less than zero indicated an avoided prey weight range. In order to test whether the segmented model method accurately described each carnivore’s prey weight spectrum as hypothesized; the literature was reviewed for additional descriptions of carnivore diet in Africa which were not used to develop the segmented models.
The mean proportion of kills in the preferred and accessible prey weight ranges across sites was determined for each large carnivore. We tested whether the mean proportion of kills in the preferred weight range was significantly different from that in the accessible prey weight range for each large carnivore, using a paired t-test .
The accuracy of detecting the correct number of changes in a predator’s prey weight preference increases with increasing prey species sample size (Table 1). Minimum species sample size at which all known break-points in prey preference were detected in 75% to 100% of simulations; and the mean (SD) absolute difference in prey category between known break-points and detected break-points at each minimum sample size.
Prey species weighing 15 kg or less are killed relative to their abundance (J.I. = 0.27±0.08, W = 149, n = 18, p <0.01) and prey species weighing more than 139 kg are consumed relative to their abundance (J.I.
The accessible prey weight ranges identified in this study account for, on average, more than 80% of each carnivore’s diet at test sites, with standard errors less than 4% (Fig. The accessible prey weight ranges account for a significantly greater proportion of each carnivore’s diet than do the preferred prey weight ranges at test sites (cheetah: t = 6.37, d.f.
Mean (USE) proportions of kills falling within the preferred and accessible weight ranges at test sites. The mean proportion of kills made by each of the five large African carnivores at test sites that fall within this study’s preferred (white) and accessible (gray) prey weight ranges.
It is based on statistically determined changes in prey preference, adhering to minimum species sample size requirements determined using simulations of known prey preference. The preferred prey weight ranges obtained using the objective approach generally support the previously determined weight ranges, particularly for cheetah and leopard.
Some refinements are notable: this study finds the preferred weight range of lions to be 180 kg broader than that determined by . In contrast, this study finds the preferred weight range of spotted hyena to be 78 kg narrower than that determined by , though the weight range found to be accessible to spotted hyena in this paper encompasses that found to be preferred by .
In addition, unlike the subjective method, the segmented model method identifies not only the weight range of prey preferred by each carnivore, but also the weight range of prey killed relative to their abundance, thereby identifying each carnivore’s accessible prey weight range. For all five large African carnivores, when tested across a diverse array of reserves in Africa, these accessible prey weight ranges accounted for over 80% of carnivore diet, with low variation across varied vegetation types and prey communities.
This study’s novel segmented model approach therefore provides an accurate description of each carnivore’s prey weight spectrum. Obtaining a broad-scale understanding of predator prey preference requires the use of data from multiple sites.
Determining the accuracy of prey species abundances in such data is challenging. For a prey weight range to be found to be significantly preferred or avoided, it must be so across many datasets.
In dietary studies of large carnivores, a potential bias is the under counting of small prey species in abundance measures. While we are unable to eliminate such a bias, we reiterate the argument made by that the underestimation of the population size of small species is likely to be counteracted by the under counting of the carcasses of these small species which are almost totally consumed.
Such an assumption can be tested using the objective approach we present in this study when a sufficient number of datasets of better known accuracy become available. Similarly, while predator sex and hunting group size can also influence prey preferences , consistent with the objectives of the published prey weight ranges that we refined, we investigated carnivore diet from a population perspective whereby a population at each site contains both sexes and all hunting group compositions.
Finally, it is important to interpret prey preferences of a single predator within the context of the large carnivore guild, since competition between predators can be an important determinant of prey preference . This study provides insights into broad-scale trends in prey preferences, and does not account explicitly for intra-guild competition in the model.
Since the majority of studies utilized in this analysis were from areas comprising complete or near-complete carnivore guilds, care must be taken when extrapolating the results of this study to areas lacking intact large carnivore guilds. As such, these broad scale preference ranges can be used as a departure point from which to test predictions regarding competitive interactions between the five large predators, with further refinements made to our understanding of preference as additional studies are conducted.
Of the five large African carnivores, only spotted hyena did not display this avoidance for prey above a certain mass-threshold. Spotted hyena are known to scavenge, and therefore may consume prey larger than that which they could physically capture , .
As the spotted hyena diet data used in this study were obtained from studies including scat analyses , scavenged prey will be included in diet descriptions and this most likely explains the spotted hyena’s lack of avoidance of large-bodied prey. Given the shortage of spotted hyena kill and prey abundance data available (21 studies amounting to 30 prey species), we were unable to exclude studies that utilized scats as the means of diet determination, in order to ensure a sufficient prey species sample size (Table 1).
Carcasses available for scavenging will generally be those large enough not to be rapidly and entirely consumed by other carnivores. This is unlikely to be an issue for the prey weight preference findings of cheetah and wild dog, who rarely scavenge , , but scavenging has been noted to comprise a proportion of the diet of lion in East Africa (16% of the diet according to ; see also , ).
Such a threshold was evident for all carnivores except leopard, who hunt prey as small as birds and rodents . This is in contrast to predators who live and hunt, at least some time, in groups (e.g. lion, spotted hyena, wild dog and male cheetah), and therefore display increased collective energy requirements which necessitate predation on larger prey , , , .
The solitary nature of leopard may also explain why this carnivore has the narrowest accessible prey weight range of the large African carnivores, despite having a similar body mass to cheetah and a larger body mass than wild dog . Curatorial predators (e.g. cheetah and wild dog) rely on speed/stamina, as opposed too stealth to hunt and therefore may not be efficient at hunting very small animals whose anti-predator response is dependent on their ability to detect a predator before it detects them .
While this is supported for wild dog in the Kruger National Park, where Stevenson flight frequently failed to elicit chases from wild dog , exceptions can arise when small prey occur in high densities which increase encounter rate and thus reduce the effort of actively pursuing small prey (e.g. ). Furthermore, an improved understanding of prey preference increases the accuracy with which we can predict, and therefore manage, the impact of predation on threatened species.
For example, roan antelope Hippocrates equines are within the weight range preferred by lion, but are usually protected from predation through their scarcity . Management actions in the Kruger National Park resulted in increased habitat congruence between roan and the lion’s preferred prey species, causing roan population collapse as a result of increased predation .
This could be done by increasing the relative abundance of other preferred prey species to provide a buffer for the Cape mountain zebra, and by managing lion numbers carefully . The prey preferences of the five large African carnivores have been used to develop predator-prey abundance models, intended to serve as a tool for estimating sustainable numbers for reintroduced carnivores .
Estimates of sustainable carnivore population sizes based solely on preferred prey may therefore underestimate sustainable predator density in areas where preferred prey species are low in abundance. While this remains to be adequately tested, the estimate for a sustainable leopard population in a mountainous area where preferred prey species are scarce was 15% lower than the observed leopard density .
Since the accessible prey weight ranges identified in this study accurately and consistently account for a large proportion of carnivore diet, the abundance of prey in these weight ranges may prove to be an improved correlate of predator density, and therefore provide improved predictions of sustainable predator population numbers. However, an 85% chance of detecting a second change would require considerably more (n = 21) prey species.
This method and the sample size guidelines could be used to refine the subjective preferred prey weight range described for tiger Panther Tigris , as well as for calculating the tiger’s accessible prey weight range. The method has been used for snow leopard Panther uncial , and data are currently being collected for multi-site prey preference analyses of dole Con alpines, jaguar Panther once, coyote Cans la trans, puma concolor, Eurasian lynx Lynx lynx and clouded leopard Neophilia nebula (Hayward peers comm).
The sources, location and timing of African studies, and the number of kills recorded in each, used to assess the mean percentage of the diet accounted for by the preferred and accessible prey weight ranges determined in this study. We thank C. Bis sett, T. Burke, M. Grover, E. Larson, J. O’Brien, S. Razz and R. Slater for carnivore kill data, and Sam Williams and an anonymous reviewer for valuable comments on an earlier version of this paper.
In: Gentleman JR, Funk SM, Macdonald DW, Wayne OK, editors. Carbone C, Gentleman JR (2002) A common rule for the scaling of carnivore density.
Kara nth KU, Nichols JD, Kumar NS, Link WA, Hines JE (2004) Tigers and their prey: Predicting carnivore densities from prey abundance. Hayward MW, Kelley GIH (2005) Prey preferences of the lion (Panther Leo).
Hayward MW, O’Brien J, Homer M, Kelley GIH (2006) Prey preference of the African wild dog Lyon ictus (Candidate: Carnivora): Ecological requirements for conservation. Jacobs J (1974) Quantitative measurements of food selection: a modification of the Forage Ratio and Inlet’s Selectivity Index.
Kirchhoff IR, Hungarian SR, Accordingly J, Rubinstein DI (2007) Habitat use and movements of plains zebra (Equus Purcell) in response to predation danger from lions. Wentworth JC, Gambling CJ, Kelley GIH (2011) Evidence for prey selection by spotted hyena in the Eastern Cape, South Africa.
Dale rum F, Homers MJ, Tunnel KE, Cameron EZ (2008) The potential for large carnivores to act as biodiversity surrogates in Southern Africa. Deena V, Sheila UV, Shelley R, Kathak B (2011) Implications of diet composition of Asiatic lions for their conservation.
Owen-Smith N, Mills MEL (2008) Predator-prey size relationships in an African large-mammal food web. Jews E, Hayward MW, Pitman RT, Swanepoel LH (2013) Effect of prey mass and selection on predator carrying capacity estimates.
Hudson DJ (1966) Fitting Segmented Curves Whose Join Points Have to Be Estimated. Mugged V (2008) Segmented: an R package to fit regression models with broken-line relationships.
Seventeen H, Berglund M (2010) An Assessment of the Density of a Large Carnivore using a Non-Invasive Method Adapted for Pilot Studies. Hayward MW (2011) Scarcity in the prey community yields anti-predator benefits.
Karloff FIT, Du Tort JT (2004) Large predators and their prey in a southern African savanna: a predator’s size determines its prey size range. Garrett RA, Brakeman JE, Becker MS, Kalinowski ST, White PJ (2007) Evaluating prey switching in wolf-ungulate systems.
Di Sylvester I, Novella O, Belgian G (2000) Feeding habits of the spotted hyena in the Niccolo Kobe National Park, Senegal. Car TM (1994) Cheetah of the Serengeti Plains: Group living in an asocial species.
Function PJ, Mills MEL, Biggs HC (2001) Factors affecting the hunting success of male and female lions in the Kruger National Park. Function PJ, Mills MEL, Biggs HC, Richardson PARK (1998) Hunting by male lions: ecological influences and sociological implications.
Skinner JD, Chiba CT (2005) The Mammals of the Southern African Subregion. Barman PJ (1974) The Social Organization of Antelope in Relation to Their Ecology.
Reich A (1981) The behavior and ecology of the African wild dog (Lyon ictus) in the Kruger National Park. Woodruff R, Lindsey PA, Romance SS, Ole Ranch SMK (2007) African wild dogs (Lyon Ictus) can subsist on small prey: Implications for conservation.
Hayward MW, O’Brien J, Kelley GIH (2007) Carrying capacity of large African predators: Predictions and tests. Harrington R, Owen-Smith N, Villon PC, Biggs HC, Mason DR, et al. (1999) Establishing the causes of the roan antelope decline in the Kruger National Park, South Africa.
Low J, Function P, Green A, Shoppers H (2012) The Applicability of Lion Prey Selection Models to Smaller Game Reserves in South Africa. Hayward MW, Jdrzejewski W, Jêdrzejewska B (2012) Prey preferences of the tiger Panther Tigris.
Lyndon S, Kshatriya S, Goal SP, Clements H, Hayward MW, et al. (2014) Prey Preferences of the Snow Leopard (Panther uncial): Regional Diet Specificity Holds Global Significance for Conservation.