Data collection in the KNP

We radio-tracked seven adult cheetahs during 1987–1990 in the southeastern region of the KNP (Broomhall 2001). These were a three-male coalition (M3), a single male (Ml), a two-male coalition (M2), and four females with or without cubs (FI, F2, F3 and F4). Three types of data collection were used, based on the duration of the observation period (Mills 1992): 1) radio-location observations, which included recording the position of the animal, the habitat it was in, and kill data where applicable; 2) short-term continuous observations, when radio-collared cheetahs were followed by vehicle for periods of 2–15 hours; and 3) three long-term continuous observations, when cheetahs were followed more or less continuously for 14 days each (two periods for M3 and one for F1). All methods of data collection have biases, but where possible direct observations of animals have obvious advantages. However, following animals in a vehicle particularly, in wooded environments, needs to be carefully executed, as the observers have to keep relatively close to the animal in order to maintain visual contact. We are unable to measure the influence this may have had on our results, but we were always careful to be as unobtrusive as possible and to hold back when the cheetahs located prey. As the observations were mainly made in daylight and the animals were radio collared it was possible to do this without missing any kills. The study area has a very high tourism impact so that most of the animals, both predator and prey, are well habituated to vehicles.

All three types of data were collected for M3 and F1, data types 1) and 2) for F2 and data type 1) only for the remaining animals. The following parameters for data types 2) and 3), were recorded when cheetahs encountered potential prey: habitat, prey species, sex and approximate age of prey, chase distance (estimated using the vehicle's odometer), kill retention time (i.e. length of time spent at the carcass, including resting periods at the carcass), and whether the carcass was appropriated by other predators (i.e. kleptoparasitism). Kills were observed until the cheetah left the carcass or the kill was kleptoparasitised. Age of prey was estimated according to tooth eruption; juveniles had erupting teeth and adults full permanent dentition.

Prey encounters were classified as: 1) a kill, 2) a failure, where cheetahs either stalked (i.e. spent at least three minutes intently watching the prey from a concealed position, and/or moved at least 5 m closer to the prey while seemingly trying not to be detected by the prey), or moved towards the prey at a faster than normal walking speed, including fast chases, but the prey escaped, or 3) no attempt, where the cheetah detected prey, but did not attempt to hunt in the manner described above. Habitat was recorded at the landscape level (Gertenbach 1983), as 1) open tree savanna plains with a moderate shrub and dense grass layer, 2) dense Acacia thickets and 3) undulating to steep Lebombo Hills with dense to moderate bush.

Analysis of KNP data

Kill data collected through radio locations and direct observations were combined for analyses as Mills (1992) and Hunter (1998) found no significant differences between these two methods in terms of bias for kills. To analyse for differences between male and female cheetah diets, prey items were separated into small (< 18 kg), medium (18–65 kg), and large (> 65 kg) weight classes, at two levels: 1) the estimated weight of the prey item according to the species, sex and age, and 2) the average weight of adult male and female animals of the prey species. Prey weights were obtained from Meissner (1982) and Owen-Smith (1988). Relative availability of male and female impala Aepyceros melampus (the dominant prey) was calculated using the average sex ratio recorded for impala (1.68 females/male) between 1986 and 1989 in the KNP (Mason 1990).

Average kill rates were determined using the three 14-day continuous observation periods only as Mills (1992) found that long-term direct observation periods were least likely to inflate kill frequency. To convert kill rate into kilogrammes of meat consumed/cheetah/day, the total weight of the prey items killed by the cheetahs were estimated using weight values for the different species according to sex and ages obtained from Meissner (1982) and Smithers (1983). Blumenschine & Caro's (1986) estimated weight of flesh of an eviscerated adult Thomson's gazelle Gazella thomsonii carcass agreed with Schaller's (1972) estimate that cheetahs consume 60% of the animal. To facilitate comparison, therefore, it was assumed that approximately 60% of the total weight of an adult impala, 65% of juvenile prey (Blumenschine & Caro 1986) and 90% of very small prey, in this case a scrub hare Lepus saxatilis, was edible.

Frequency data for hunting and killing across habitat types were derived by combining the hunting locations of three cheetahs (M3, F1 and F2) and kill locations of four cheetahs (M3, FI, F2 and F3). For each analysis, a minimum convex polygon was drawn around the cheetahs' home ranges to determine the area (in km²) available for hunting and killing in the three habitat types using GIS Arcview (Broomhall, Mills & du Toit 2003). A chi-square goodness-of-fit test was used to determine if cheetahs were killing impala in proportion to their relative availability in the different habitat types (Hunter 1998). Relative availability was determined by using the unpublished annual aerial impala census data (Viljoen, Rochat & Wood 1994) to calculate the mean number of impala per habitat type for the study area. Impala density (animals/km²) was calculated for each habitat type. Following any significant results from the above chisquare tests, Bonferroni confidence intervals were performed to determine preference or avoidance by cheetahs of particular habitat types (Neu, Byers & Peek 1974, Byers & Steinhorst 1984).

Across-ecosystem comparisons

For a comparison of prey composition (prey size and age) across different savanna ecosystems, suitable data were synthesised from 10 different studies in southern and East Africa. Study sites were as follows: East Africa (Graham 1966), a broad survey conducted across Uganda, Tanzania and Kenya, Serengeti National Park (SNP; Kruuk & Turner 1967), Kafue National Park (Kafue NP; Mitchell, Shenton & Uys 1965), Matusadona National Park (MNP; Zank 1995), the Kgalagadi Transfrontier Park (KTP; Mills 1984), Suikerbosrand Nature Reserve (SBNR; Pettifer 1981a), Phinda Resource Reserve (PRR; Hunter 1998), Mala Mala Game Reserve (MM; Radloff & du Toit 2004); Timbavati & Klaserie Private Nature Reserves (TNR; Pettifer 1981b), and KNP (this study).

For each study site, prey was divided into adults and juveniles for small, medium and large prey species as above. Laurenson (1995b) classified medium-sized prey as ranging within 15–60 kg, so this was adjusted slightly to facilitate analysis in this study. Prey weights were obtained from Owen-Smith (1988) where the average weight of adult males and females determined the size category of the prey. The adults and juveniles of small prey were combined because studies often did not classify small prey items in this manner; particularly when considering prey items such as birds, hares, rodents and small carnivores. Studies with no reported kills of small prey were excluded from the analysis as small prey are known to form a significant part of the cheetah's diet (Labuschagne 1979), but are often underrepresented in studies due to the method of data collection used by the researcher (Mills 1992).

Table 1.

Habitat description in selected cheetah study sites across southern and East Africa. Each study site was ranked subjectively for cover availability, where the site with the least amount of cover was given a value of one; see Fig. 1.

Further comparisons across ecosystems were conducted concerning hunting success, chase distance, kleptoparasitism, kill retention time and kill rates. Study sites for these analyses were SNP (Schaller 1972, Caro 1994), Nairobi National Park (NNP; Eaton 1970, McLaughlin 1970 cited by Schaller 1972), KTP (Labuschagne 1979), MM (G.T. Radloff, unpubl. data), PRR (Hunter 1998), TNR (Pettifer 1981b), Suikerkop Nature Reserve (SNR; Pettifer 1981b), and KNP (this study). The relationships between cover availability and various parameters pertaining to feeding ecology, i.e. chase distance, hunting success and kleptoparasitism, were explored. For these analyses, study areas were ranked according to cover availability (open to closed cover), where the area with the least amount of cover was given a value of one (Table 1). Kleptoparasitism values were not standardised to control for variation in predator density because the SNP and KNP had similar total densities of lion Panthera leo and spotted hyaena Crocuta crocuta (Stander 1991), and it was assumed that MM, which adjoins the KNP and is unfenced, had the same predator density as the KNP.

Cheetah in the KNP

Of the nine prey species observed to be killed by cheetahs in the southeastern KNP, impala were the most frequent for both males and females (Table 2). For all prey species, cheetahs took more juveniles (61%) than adults (39%), particularly of large prey species, although male cheetahs took impala adults more frequently than juveniles. Of the 18 adult impala killed, 78% were male and 22% were female. Cheetahs preyed on male and female impala at significantly higher and lower frequencies, respectively, than their availability would predict (x² = 12.7, df = 1, P < 0.001).

Table 2.

Cheetah prey composition in the Kruger National Park.

Table 3.

Cheetah hunting behaviour and density of impala in different habitat types in the southeastern region of the Kruger National Park.

There was a difference in the way in which male and female cheetahs utilised prey of different weight and size classes. The male cheetahs' diet consisted of a greater proportion of larger prey items (21, 68 and 12% large-medium-, and small-weighted prey, respectively), while female cheetahs caught smaller prey items (44 and 56% of medium- and small-weighted prey, respectively). When comparing male and female cheetah selection of prey based on average adult size, there was a significant difference (x² = 18.3, df = 2, P < 0.0001). The males' diet consisted of larger prey species, such as kudu Tragelaphus strepsiceros, waterbuck Kobus ellipsiprymnus and zebra Equus burchelli, whereas the females caught common duiker Sylvicapra grimmia and steenbok Raphicerus campestris.

Although the hunting success for the three male coalition (M3) was higher (25%) than for the two female groups (16 and 17%, respectively) this was not significant (x² = 2.32, df = 1, P > 0.05). However, the frequency of hunting attempts per prey encounter (45 vs 69 and 71%, respectively) was significantly higher in the female groups (x² = 5.758, df = 1, P < 0.05). Kill rates for M3 were 1 kill/7 days (or 1.4 kg meat/cheetah/day) and 1 kill/3.5 days (or 1.4 kg meat/cheetah/day) for two 14-day continuous observation periods. Although F1 accompanied by two large cubs was followed for one 14-day continuous observation period, the kill and consumption rates were not accurately determined as contact was lost with these animals for 17 hours. However, judging by their stomach sizes when relocated they had obviously not made a big kill, although a small kill may have been made. At least two kills were made during the 14-day period, giving a minimum kill rate of 1 kill/4.6 days. One of the kills was kleptoparasitised by a spotted hyaena, so the minimum amount of meat consumed by the three animals was approximately 0.4 kg meat/cheetah/day. Using the pooled data for males and females, a significant difference was found between the mean chase distance of successful ( = 189 m, SE = 22.9, N = 26) versus unsuccessful ( = 96 m, SE = 9.41, N = 89) hunts (t = 4.36, df = 113, P < 0.0001). The mean kill retention time was 165 minutes (SE = 59, N = 9). When combining data of males and females, kleptoparasitism was 12% (N = 34).

In Table 3 aspects of the cheetahs' hunting behaviour in the three habitat types are analysed. Once prey was detected, cheetahs attempted more hunts per prey encounter in the open savanna than in Acacia thickets and the Lebombo Hills, with significantly more hunts than expected in the open savanna and less than expected in the Acacia thickets (x² = 153, df = 1, P < 0.0001). Hunting success (kills/hunting attempt) was also greater, but not significantly so, in the open savanna than in the Acacia thickets. The frequency of kills per habitat type was significantly different from the expected based on habitat available for killing within the cheetahs' home ranges (x² = 11.3, df = 2, P < 0.01), most kills being made in the open savanna. The frequency of hunting attempts per habitat type also differed significantly from the expected based on habitat available for hunting within the cheetahs' home ranges (x² = 14.9, df = 2, P < 0.001), with most hunting attempts also occurring in the open savanna. Of the three available habitat types, Bonferroni confidence limits indicated that the open savanna was used significantly more and the Lebombo Hills significantly less than expected for killing and hunting. The Acacia thickets were used significantly less than expected for killing but were used in proportion to availability for hunting (Table 4). Impala were killed at significantly different frequencies to those predicted based on their occurrence across different habitat types (X² = 22.5, df = 2, P < 0.0001), with more being killed in the open savanna. Bonferroni confidence limits indicated that cheetah killed significantly more impala in the open savanna and significantly less impala in the Lebombo Hills than expected, while they were killed in proportion to their availability in the Acacia thickets (see Table 4). The densities of impala were higher in the Lebombo Hills and Acacia thickets than the open savanna.

Table 4.

Habitat selection by cheetahs for killing, hunting and hunting impala in the southeastern region of the Kruger National Park. The symbols indicate whether use was significantly greater (+), less (-) or no different (0) to the expected use based on the proportion of habitat available within the cheetahs' home ranges for killing and hunting, and the proportion of impala available in the different habitat types. The level of significance was set at 0.05.

Table 5.

Proportions (in %) of size categories and age classes of cheetah prey in 10 study sites across southern and East Africa. The size categories include: small- (<18 kg), medium- (18–65 kg) and large-sized prey (> 65 kg).

Across-ecosystem comparisons

In 10 study sites across southern and East Africa, the adults of medium-sized prey (18–65 kg) occurred most frequently in the cheetah's diet, followed by the juveniles of medium- and large-sized prey (Table 5). There was, however, a significant variation in the size and age groups of prey taken by cheetahs across study sites (ANOVA: F = 7.406, df = 49, P < 0.0001). In the KNP, impala were the most abundant prey and the most common species in the cheetah's diet (see Table 2). On the SNP, it was Thomson's gazelle (Kruuk & Turner 1967, see also Caro 1994) and in KTP springbok Antidorcas marsupialis (Mills 1984). All these species are in the medium-sized prey category.

In Kafue NP, SBNR and PRR, cheetahs selected a large proportion of adults in the large-size category (> 65 kg), namely puku Kobus vardoni, blesbok Damaliscus pygargus and nyala Tragelaphus angasi, respectively (Mitchell et al. 1965, Pettifer 1981a, Hunter 1998), compared to others areas. The average weight of male and female puku, however, is 67 kg, and therefore borders close to the division between medium- and large-sized prey. In SBNR there was a preferred selection for blesbok females (60 kg) and juveniles, which Pettifer (1981a) explained may have been due to males weighing up to 80 kg. Hunter (1998) suggested that the behaviour of nyala (average weight: 85 kg) browsing in more open areas near cover and their sluggish nature made them more vulnerable to cheetah predation.

In the SNP, Kafue NP, PRR, KNP and TNR cheetahs utilised a greater proportion of juveniles of large prey, and in KNP and MM cheetahs took a greater proportion of small-sized prey compared to other areas.

Relationships between the rank of cover per study site and chase distance in successful hunts, percent hunting success and percent kleptoparasitism are shown in Figure 1 and Table 6. This small data set suggests that chase distances, hunting success and amount of klep toparasitism are higher in more open areas, but that the differences between habitats are not great. Study sites with the least cover had the longest mean chase distance, while those with greatest cover had the shortest mean chase distance. All study sites had longer chase distances for successful hunts (see Table 6). No patterns were found in mean kill retention time and kill rate across study sites (see Table 6).

Figure 1.

Relationship between rank of cover at a study site and mean chase distance (A; in m) in successful hunts, hunting success (B; in %), and kleptoparasitism (C; in %). See Table 1 for key to protected area initials and brief descriptions of the habitat in each area.

Table 6.

Aspects of cheetah hunting behaviour and incidents of kleptoparasitism in eight study sites across southern and East Africa. ** indicate that no data are available.

Prey selection

Diet preferences of cheetahs in different areas reflect differences in prey species and their abundance (Stander 1991, Caro 1994, Mills 1998; see Table 5). The cheetah's main food, however, is medium-sized prey, which represents an average of 60% of the diet across ecosystems (see Table 5). The juveniles of large-sized prey also form an important part of the diet in many areas, most of these fall into the medium-sized prey category (18–65 kg).

In the KNP study, smaller prey formed a significantly more important part of the female cheetah group's diets than of the three-male cheetah coalition (see Table 2). This is at least partially explained by the fact that the size and composition of the hunting group may affect prey size and species (Eaton 1974, McVittie 1979, Caro 1994). In PRR, Hunter (1998) found that male cheetah coalitions killed mostly male nyala (120–130 kg) while female cheetahs killed mostly female nyala (60–70 kg). In the SNP, Caro (1994) also found that larger groups of cheetah hunted wildebeest Connochaetes taurinus more often than did smaller groups. Therefore, the group size and sex of the animals observed may influence the results of prey selection studies.

Small prey represented an average of 12% in the cheetah's diet across the African savanna study sites (see Table 5). Small prey, however, are usually underrepresented (Stander 1991, Mills 1992) because studies often depend on data from carcass remains (Pienaar 1969, Pettifer 1981a) or opportunistic observations (Mitchell et al. 1965). Kills of small prey are usually unobserved because of the rapid consumption time and lack of remains (Mills 1992). The large percentage of small prey recorded in the cheetah's diet in this study (27%) and in Radloff & du Toit's (2004) study in Mala Mala (24%) are probably more representative of the proportion of small prey because these studies relied mainly on continuous observation data (see Methods). Alternatively, these are the areas with the highest cover values (see Table 1), and so the habitat structure may also account for the higher kill rate of smaller prey.

Preferences by cheetahs for male impala in the KNP study are paralleled by cheetahs' preferences in SNP for male Thomson's gazelle (Fitzgibbon 1990) and in KTP for male springbok (Mills 1990). Fitzgibbon (1990) describes how male gazelle are more vulnerable than females and preferentially selected by cheetahs because they tend to occur on the periphery of groups, have greater nearest-neighbour distances, are less vigilant and are found in smaller groups than females. The same conditions may apply to male impala and springbok as they show similar social structures and behaviour (Estes 1991). As males are more expendable than females, this greater vulnerability of males to predation has the effect of lessening the impact of predation on prey populations.

Kill rate and consumption

Group size, presence of cubs, prey size and availability, habitat structure and competition with other predators affect kill rates (Pettifer 1981a, Caro 1994, Durant 2000, this study). These probably explain the large variations found in cheetah kill rate across African savanna ecosystems (see Table 6). The rate of food consumption needed to keep a cheetah in healthy condition in a zoological garden is 1.3–1.8 kg/day (Crandall 1964). Although the calorific needs of zoo animals are lower than those of wild cheetahs, captive animals are usually fatter; therefore the male cheetah coalition in the KNP appeared to be obtaining an adequate diet (1.4 kg/cheetah/day). The female cheetah's (F1) consumption rate, with two large cubs, was considerably lower (0.4 kg/day), although as explained this figure may be slightly lower than the actual figure. However, Caro (1994) estimated that cheetah mothers with old offspring ate as little as 0.5 kg/day because of direct competition from their large cubs.

Kill retention time

Kill retention time may be affected by group size, prey size, predator densities, knowledge of competing predators (Schaller 1972, Pettifer 1981b, Hunter 1998), or amount of available cover. These may explain the large variation found in kill retention times across ecosystems (see Table 6). In SNR, TNR and PRR, cheetahs were acquired from captive-breeding programmes (Pettifer 1981b) or Namibia (Hunter 1998) for re-introductions, and had not been subjected to competition from other large predators. Pettifer (198 lb) discussed this as the reason for the exceptionally long kill retention times in SNR and TNR. Hunter (1998) attributed the lack of direct competition in PRR to the long hours cheetahs spent at carcasses. Differences in kill retention time between the SNP and KNP, with similar densities of competing predators (Stander 1991) were similar, although differences in availability of cover would have predicted shorter retention times in more open habitat.

Hunting and habitat

Cheetahs initiated more hunts and killed more frequently in the open savanna of the KNP than in other available habitats with thicker bush (see Table 3). It is important to note, however, that KNP open savanna is not like the open grassland plains of the Serengeti. It is an open woodland with a moderate shrub layer and tall grass throughout. The preference by cheetahs for the most open habitat in the KNP for hunting (see Table 4) is particularly evident when considering that the cheetah's main prey (impala) occurred at greater densities in the Acacia thickets and Lebombo Hills (see Table 3), yet were hunted and killed significantly more in the open savanna (see Table 4). In PRR, which consists of overlapping open to closed bushveld habitat, cheetahs also preferred the open grasslands for hunting (Hunter 1998) and in MNP, cheetahs used the open foreshore grass land predominantly for hunting (Purchase & du Toit 2000).

Chase distance and hunting success

Chase distances and hunting success are difficult to measure (Mills 1990), and comparisons between studies are difficult to make. Nevertheless, chase distances appear to be shorter in more wooded habitats, the average for KNP was 2.3 times less than the average for SNP (Prediction I; see Fig. 1A and Table 1). Caro's (1986) study on the SNP and Eaton's (1974) study in Nairobi National Park found that cheetahs were more likely to be successful at hunting when they were able to get closer to their prey before rushing. Compared to the KTP and SNP, cheetahs in the KNP had significantly longer chase distances in successful (189 m) than in unsuccessful hunts (96 m; see Table 6). The success of longer chase distances indicates that cheetahs are able to quickly gauge their chances of success in a chase and give-up early if failure is predicted.

Hunting success rates between studies did not vary greatly but appeared to be higher in more open habitats (Prediction ii; see Fig. 1B and Table 1). In the KNP, cheetahs also appeared to have a greater hunting success and hunted more often in the open savanna habitat (see Tables 3 & 4). Therefore, greater tree and shrub cover in woodland habitats may obstruct the cheetah's high-speed hunting strategy, thereby lowering hunting success.

Kleptoparasitism

Across African savanna ecosystems, cheetahs appeared to be kleptoparasitised less in more wooded habitats (Prediction iii; see Fig. 1C and Table 1). Paulson (1985) considered four effects that open habitat has on host and parasite, three of which are relevant to cheetahs: 1) kleptoparasites can observe and follow hosts more easily, 2) kleptoparasites can observe prey capture and carrying, and 3) hosts are less able to hide from kleptoparasites. Considering these effects, cheetahs in an open grassland ecosystem like the SNP, with a short to medium grass layer, are expected to be more vulnerable to kleptoparasitism than in areas like the KNP (Myers 1975).