Facilitation of a free-roaming apex predator in working lands: evaluating factors that influence leopard spatial dynamics and prey availability in a South African biodiversity hotspot

Ethical statement

Relevant permissions to conduct our research were granted by the Social, Behavioural and Education Research Ethics Committee at Stellenbosch University (Project ID #15315), CapeNature (Permit #CN44-59-12321), and in writing by all landowners involved. Data collection was performed using camera-traps, which is a non-invasive research method. All data, including any images captured of human subjects, were treated as strictly confidential.

Study area

The Piketberg region (hereafter Piketberg) encompasses a transformed landscape that is primarily characterised by mixed agricultural practices. Our study area is located approximately 130 to 160 km north of Cape Town, South Africa, and situated southwest of the Cederberg region (Fig. 1). The survey area was approximately 1,500 km² in size (53–864 m above sea level), covering 55 privately owned mixed agricultural farms extending north from the town of Piketberg to Paleisheuwel, with Citrusdal in the east and Aurora on the western boundary (Fig. 1). The area consists of natural vegetation forming a mosaic, highly fragmented by livestock (e.g. cattle (Bos taurus, Linnaeus, 1758), horses (Equus ferus caballus, Linnaeus, 1758), sheep (Ovis aries, Linnaeus, 1758), goats (Capra hircus, Linnaeus, 1758), pigs (Sus domesticus, Erxleben, 1777)), fruit, and other crop farmlands in and around mountainous terrain (Linder, 1976; Mucina & Rutherford, 2006).

The Cederberg region (hereafter Cederberg), known for its rugged remoteness, is a largely protected area first proclaimed in 1973 and located 200 to 250 km northeast of Cape Town (Fig. 1). This survey area was approximately 1,700 km² in size (254–1,455 m above sea level) and included the Matjiesrivier Nature Reserve and the Cederberg Wilderness Area. Both areas are formally protected and managed by the provincial conservation regulation body and authority in the Western Cape, known as CapeNature (Fig. 1). Privately owned areas included were Bushmans Kloof Wilderness Reserve, community owned land used for the harvesting of rooibos and limited pastoralism by subsistence farmers, and the Cederberg Conservancy, consisting of pro-conservation farms that are used for ecotourism and largely kept in a natural ecological state. The two main biomes present are Fynbos and Succulent Karoo in mountainous terrain (Mucina & Rutherford, 2006).

The Western Cape has a Mediterranean type climate characterised by hot, dry summers and cold, wet winters (Cowling & Holmes, 1992). Localised climatic conditions vary across the greater study area because of its mountainous nature, and the vastness of the landscape makes it extremely difficult to describe the climate of the study area in general terms. In essence, average annual rainfall appears slightly higher, and average summer air temperatures somewhat lower in Piketberg compared to the Cederberg (Climate-Data.org, 2020). Average winter air temperatures are more uniform across both regions, but snowfall is more abundant in the Cederberg which is generally at a higher altitude (Climate-Data.org, 2020).

Field sampling methods

Data analyses

Species richness

Piketberg had an estimated potential prey species richness (S = 26) similar to that of the Cederberg (S = 27). The rarefied species accumulation curves do however show more species detected per unit effort in the Cederberg compared to Piketberg (Fig. 2). Both the Shannon-Wiener (H′) and Simpson diversity (Ds) indices were only marginally higher in the Cederberg (H′ = 2.64, Ds = 0.90) compared to Piketberg (H′ = 2.46, Ds = 0.89). Thirty potential natural prey species were photographed across the two regions (Cederberg: 27; Piketberg: 25), with five species unique to the Cederberg and three to Piketberg (Table S2).

Community structure

Differences in community structure were largely complementary, providing a similar pattern across trophic guilds for both the Cederberg and Piketberg (Fig. 3). The same number of carnivore species were detected in both regions (n = 10), but no Cape fox (Vulpes chama, Smith, 1833) or large grey mongoose (Herpestes ichneumon, Linnaeus, 1758) were recorded in the Cederberg, whereas black-backed jackal and water mongoose (Atilax paludinosus, Cuvier, 1826) were not detected in Piketberg. Herbivores were the most frequently encountered guild across both regions; more herbivorous species were encountered in the Cederberg (n = 13) than in Piketberg (n = 11). Main leopard prey species (i.e., baboon, duiker, grysbok, hyrax, klipspringer, porcupine, and rhebok) were all accounted for in both communities (Table S2).

Across all guilds, global trap rates (RAIs) for species detected were generally higher in Piketberg (Fig. 3 & Table S2). Noteworthy exceptions included hyrax (Cederberg: 4.05; Piketberg: 3.16), klipspringer (Cederberg: 3.56; Piketberg: 2.41), and rhebok (Cederberg: 0.84; Piketberg: 0.14). A marked difference was the higher RAIs of all carnivores >4 kg—including leopard (Cederberg: 3.52; Piketberg: 4.65)—as well as most medium (>0.5 kg and <100 kg) herbivores, in the Piketberg region (Fig. 3 & Table S2). Baboons were the most frequently detected species across both communities (Cederberg: 608 events; Piketberg: 592 events), followed by hyrax in the Cederberg (410 events) and porcupine in Piketberg (540 events). No single species dominated (i.e., >50% of total trap rates) in any of the two communities.

High levels of similarity between the two study regions were shown by both the Jaccard’s Index (J = 0.73) and Sørensen’s Coefficient (CC = 0.85). The Jaccard’s Index suggested that 73% similarity exists between the contrasting communities. Similarly, the Sørensen Coefficient suggested that the ecological distance that separate the two regions is merely 15%, supporting close relatedness of ecological make-up in each community.

Spatial behaviour

Camera-trap stations were excluded from occupancy (i.e., space-use) analyses when cameras were operational for <80% of occasions (see Fig. 1). Baboon and rhebok were excluded due to insufficient model fit (Table S3). Only strong relationships with space-use and detection probabilities are discussed (see Table 1). In essence, factors that strongly influenced the spatial dynamics of leopards and their main prey species in the Cederberg were primarily environmental (56%), whereas anthropogenic (32%) and biotic variables (12%) played a less significant role (Table 1). In contrast, anthropogenic variables (41%) dominated in Piketberg relative to notable influences by biotic (36%) and environmental (23%) factors (Table 1).

Leopard

Leopards occurred across the landscape in both communities and were detected at 55 camera-trap stations in Piketberg and 60 stations in the Cederberg, providing naïve occupancy estimates of 0.852 (Piketberg) and 0.833 (Cederberg) respectively. The spatial dynamics of leopards differed between the Cederberg and Piketberg, except for both leopard populations preferring to utilise areas further away from human habitations (Table 1; Figs. 4A and 5A). In the Cederberg, leopards also preferred areas in closer proximity to permanent rather than seasonal water sources (Table 1; Fig. 4B) but were more likely to be detected further away from water (Table 1; Fig. 4E). Furthermore, leopards of the Cederberg were generally less likely to be detected in areas where sympatric meso-carnivores were more abundant (Table 1; Fig. 4C) and at higher altitudes (Table 1; Fig. 4D). In Piketberg, leopards were less likely to be detected if present at sites characterised by Sandveld vegetation and in areas utilised by livestock (Table 1; Figs. 5B–5E).

Prevalence of leopards

Contrary to expectations, our results, in terms of both relative abundance indices (RAI: Cederberg: 3.52; Piketberg: 4.65) and naïve occupancy (ψ_naïve: Cederberg: 0.833; Piketberg: 0.852), suggest that leopards are at least as relatively widespread across the agriculturally transformed Piketberg landscape, and potentially even more abundant, than they are in the primarily protected Cederberg region. We similarly observed greater relative abundances of all carnivorous species >4 kg shared between the two communities (Fig. 3 & Table S2). Assuming comparable average activity levels, leopards of Piketberg may therefore occupy smaller or more overlapping home ranges than leopards found in the Cederberg (refer to Müller et al., 2022b). Recently, Snider et al. (2021) showed that it is indeed common for free-roaming leopard home-range size to be smaller (inferring greater density) within areas of greater human population density. Alternatively, should activity levels differ greatly between the two subpopulations, leopards could also occupy larger home ranges in Piketberg (Neilson et al., 2018; Rogan et al., 2019). Since the relationship between occupancy (or space-use), abundance, and density is influenced by the number of individuals, home range size, and the degree of spatial overlap between individuals, neither relative abundance nor the space-use parameter can be used to ascertain differences in density with absolute certainty (Rogan et al., 2019). Nonetheless, greater predator abundances can only be sustained by a greater relative carrying capacity (i.e., prey availability). Higher RAIs were also noted in Piketberg compared to the Cederberg for most medium sized (>0.5 kg and <100 kg) herbivores (Table S2), which are regarded as the primary prey component for leopards (Hayward et al., 2006). Therefore, prey populations in Piketberg appeared sufficient to facilitate and sustain relatively greater, potentially denser, and at least as equally successful predator populations to those found in the Cederberg. Notable differences that we observed in the RAIs of main leopard prey species between the two study regions, however, suggests that leopard diet composition is likely to differ (based on availability; Hayward et al., 2006) in the agriculturally transformed landscape.

The main prey species considered in this study are known to comprise approximately 85% of the biomass consumed by leopards in the Cederberg (Müller et al., 2022a), while in other areas of the Western Cape they comprise as much as 91% (Boland) and as little as 57% (Little Karoo) of leopard diets (Mann et al., 2019). Grey rhebok, klipspringer, and rock hyrax were more abundant in the Cederberg, while Cape grysbok, Cape porcupine, chacma baboon, and common duiker were more abundant in Piketberg (Table S2). Leopards in the Cederberg appear heavily reliant on hyrax and klipspringer in particular, which together constitutes ~61% of biomass consumed (Martins et al., 2010; Müller et al., 2022a). Despite lower relative abundances recorded for both these prey species in Piketberg (Table S2), where leopard diet composition remains unknown, the leopard population here appeared unaffected. Our findings thus suggest the persistence of leopards in this agriculturally transformed region, highlighting the adaptability, but also the potential vulnerability, of these large apex predators. Leopards in the Soutpansberg and Waterberg mountains (Limpopo Province) of South Africa have also been shown to thrive outside of protected regions (Chase Grey, Kent & Hill, 2013; Swanepoel, Somers & Dalerum, 2015), as is the case across a densely populated agricultural landscape in India (Athreya et al., 2013). Therefore, together with our results, it is evident that a landscape level approach is needed to ensure that the appropriate conservation policies, laws, and practices are implemented (Athreya et al., 2013) to ensure the safeguard of leopards throughout the entirety of their distributional range and not just in protected areas. The importance of protected regions is, however, not to be underestimated (see Mohammadi et al., 2021).

Natural prey species availability

Our camera surveys appear to have recorded almost all natural potential prey species (mammals >0.5 kg) present, as the number of species captured was very close or equal to the total number estimated in both communities. Community composition of potential prey species did differ, but niche composition remained relatively intact across both regions, and each supports complete communities of carnivores and herbivores (Fig. 3). Generally, carnivores do not play a significant role in leopard diet in the Western Cape (Martins et al., 2010; Drouilly, Nattrass & O’Riain, 2018; Mann et al., 2019; Müller et al., 2022a), although elsewhere small carnivores are commonly killed (Palomares & Caro, 1999; Hayward et al., 2006). Herbivores unique to each community in our study (gemsbok in Cederberg; kudu in Piketberg: Table S2) were both large antelope species (>100 kg) which also generally do not constitute a major component of leopard diet in the Western Cape, except in the Little Karoo (Martins et al., 2010; Drouilly, Nattrass & O’Riain, 2018; Mann et al., 2019; Müller et al., 2022a). Here, their marked presence in leopard diet has been attributed to land-use change, characterised by an increase in game farming in recent years (Mann et al., 2019). It is however improbable that the very low abundances (Table S2) and limited distributions (detected at only one station each) we recorded for these large antelopes in Piketberg would effectively support its leopard population. Instead, predation on these introduced individuals would be likely to ultimately result in farmer-predator conflict and increase the level of risk that leopards are exposed to (Constant, Bell & Hill, 2015). Hence, the minor differences observed in potential prey species composition and richness between the Cederberg and Piketberg are unlikely to influence leopard diet substantially, and consequently population persistence, in the latter.

Variance of main prey species

Leopards in Piketberg are more likely to rely on alternative main prey species to those primarily utilised in the Cederberg, based on observed differences in their perceived availability in each community (Table S2). By considering the factors that strongly affects habitat use of main prey species (Table 1), we broadly infer some potential underlying drivers of these differences. Due to the inability to model probabilities of space-use and detection for baboon and rhebok, we did not speculate on probable causes for differences in their availability (i.e., RAIs). It should also be noted that our results for hyrax may be unintentionally skewed because of their restricted habitat (being confined to rocky outcrops; Skinner & Chimimba, 2005) not being accounted for in our camera set-up procedures; we advise caution in interpretation thereof.

Unsurprisingly, the relative impact on prey species by anthropogenic factors accompanying agricultural practices (e.g. roads, disturbances, habitations) was greater in Piketberg, but not all species were negatively influenced (Table 1). In essence, anthropogenic drivers were generally more likely to strongly influence preferred main prey species of the Cederberg (hyrax and klipspringer) in a negative manner, but alternative main prey species (duiker, grysbok, and porcupine) appeared less sensitive, some even showing a potential preference for disturbed sites. These alternative prey species are therefore likely to play a relatively larger role in leopard diet and facilitation of the leopard population in Piketberg. The greater perceived abundance of duiker, grysbok, and porcupine in the agriculturally transformed region may further be facilitated by a preference for readily available crop food resources as these species are known to frequent the fringes of agricultural land (Birss, Relton & Selier, 2016; Bragg & Child, 2016; Palmer et al., 2016). Elsewhere in the Western Cape, however, it has recently been shown that duiker and grysbok both remain dependent on natural vegetation even within severely transformed landscapes (Jansen van Vuuren, Fritz & Venter, 2022). We therefore believe that suitable natural habitat within the mosaic Piketberg landscape plays a pivotal role for sustaining these species. Thus, we promote the maintenance of natural vegetative corridors within and between transformed lands. Remaining natural vegetation in Piketberg can generally be considered denser than across the Cederberg, thereby having the potential to further cater for grysbok and porcupine which are both known, and shown by this study, to require sufficient vegetative cover (Bragg & Child, 2016; Palmer et al., 2016). In contrast, the sparser Karoo vegetation of the Cederberg appear to be favoured by hyrax and klipspringer (Birss et al., 2016; Visser & Wimberger, 2016). These two species also appeared to be the most affected by top-down influences of predators as both species in both communities seemingly avoided areas heavily utilised by caracal and black-backed jackal, resulting in an indirect spatial overlap with leopards in the Cederberg. Interestingly, leopard RAI strongly influenced main prey species in Piketberg only, exhibiting direct overlap with grysbok. Yet again, hyrax and duiker in Piketberg appeared to be influenced in an opposing manner. In the Cederberg and elsewhere, hyrax are particularly favoured as prey by both leopard and caracal (Hayward et al., 2006; Müller et al., 2022a). It is thus also plausible that hyrax may in fact experience significant predation pressure and therefore their numbers and activity may appear relatively limited in areas shared more frequently with leopards in Piketberg (Wittmer, Sinclair & Mclellan, 2005).

Livestock—an alternative food source?

Aside from alternative main prey species playing an important role, unnatural prey (i.e., livestock or domestic species) might also supplement leopard diet in Piketberg. In a national park in Pakistan and a human-dominated landscape in India for example, leopards have previously been shown to be almost completely dependent on livestock and other domestic species as prey (Shehzad et al., 2015; Athreya et al., 2016). Our decision to exclude livestock and other domestic species from abundance analyses was because the vastly greater use of livestock proof fences in the Piketberg region prevented accurate and comparable detection of livestock at camera-trap stations. While fences do not prevent the movement of leopards across a landscape, the energetic costs to leopards that accompany their presence may be a driving factor for the killing of more livestock (Wilmers et al., 2017). Müller et al. (2022a) showed that 7% biomass of leopard diet in the Cederberg was comprised of livestock. Subsequently, we believe a greater proportion can be expected in Piketberg, which is a conclusion that seems to be supported by higher levels of livestock predation events reported in recent years (C Luyt—Community Outreach Officer at the Cape Leopard Trust, C Luyt, 2020, personal communications). Although not found to be strongly correlated in either region, the inverse relationship between space-use by leopards and presence of livestock in the two contrasting study regions (Table 1), is noteworthy. Leopards in the Cederberg tended to avoid areas with signs of livestock, while in Piketberg they appear to have shown a greater preference for areas with livestock. This observation could suggest a plausible tendency for greater reliance on livestock by leopards in Piketberg. In the Cederberg, livestock roam more freely but tend to be guarded by herders, whereas in Piketberg they are generally fenced and unguarded. Unguarded, fenced livestock that are not completely predator-proofed may result in leopards being attracted to livestock as prey, especially when preferred natural prey abundances are low (Odden, Nilsen & Linnell, 2013; Khorozyan et al., 2015). This may subsequently result in an increase in illegal retaliatory killings, making the leopard population in Piketberg particularly vulnerable (Inskip & Zimmerman, 2009; Soofi et al., 2022).

Importantly, if livestock serve as regular prey, this has the potential to result in less pressure and reduced ecological regulation of natural prey species, regarded as agricultural pests (Norton, 1980; Kingdon, 1982; Estes, 1991; Skinner & Chimimba, 2005). Examples of this have been described for other apex predators like snow leopards (Panthera uncia, Schreber, 1775) in Nepal and dingoes (Canis lupis dingo, Meyer, 1793) in Australia (Johnson & Wallach, 2016; Shrestha, Aihartza & Kindlmann, 2018). Conversely, leopards that are less reliant on livestock as prey, provided sufficient preventative measures for livestock predation are employed, can partially limit population explosions of their natural prey species (O’Bryan et al., 2018). Effective preventative mitigation of conflict with crop pests (i.e., prey species) may also cater for leopards, offering sufficient natural prey availability in the future which may minimise livestock losses (Odden, Nilsen & Linnell, 2013; Khorozyan et al., 2015). Hence, leopards and their prey species can be regarded as assets on a landscape scale, and pro-active (preventative and non-lethal) measures employed together by livestock and crop farmers in an agricultural community can thus benefit them both instead of either one indirectly inducing conflict onto the other. We therefore agree with Du Toit, Cross & Valeix (2017) that a shift in attitude towards asset management, rather than problem control by means of retaliatory killings, would be advantageous as the removal of conflict-prone species is likely to be counter-productive for the community (Conradie & Piesse, 2013; Lennox et al., 2018).

Impacts on leopard behaviour

Besides potential for retaliatory killings of leopards, a further direct negative impact due to humans was reflected by the avoidance of human habitations by leopards in both study regions (Table 1; Figs. 4A and 5A). This reiterates that anthropogenic development is responsible for habitat loss (Swanepoel et al., 2013; Jacobson et al., 2016). Environmental factors that directly dictated the spatial dynamics of leopards differed between the two regions (Table 1). Greater detectability further away from water in the Cederberg, where sources are presumably more limited, can be explained since leopards are largely independent of water (Bothma, 2005). Leopards in this region did however show a preference for areas closer to permanent (e.g. rivers; Table S1) rather than seasonal (e.g. streams; Table S1) water sources, but this is likely to be a simple consequence of the seasonal water sources being dry and unusable at the time of our study. The lower detection probability of leopards in Sandveld vegetation and areas presumably used for grazing in Piketberg indicate that fine-scale habitat utilisation by leopards is likely to be selective and affected by agricultural activities. The influence of altitude observed in the Cederberg is assumed to be a consequence of the slightly higher density recorded for leopards in the region during the summer months, which relates to individuals occupying smaller home ranges on the more accessible lower mountain slopes at this time of year (Müller et al., 2022b). Finally, the relationship we observed between leopards and meso-carnivores in the Cederberg is consistent with the conclusions of Müller et al. (2022a): caracal (meso-predator) tend to avoid leopards (apex predator) in time and space. Interestingly, the availability of natural main prey species did not directly dictate spatial utilisation by leopards, which is probably because leopards (and natural prey) occurred widely across the study area. It is important to note that the impacts we observed on the behaviour of various main prey species, both environmental and anthropogenic, do have the potential to indirectly affect leopards going forward. Therefore, continuous monitoring and evaluation of the environmental and anthropogenic factors affecting the ecology of both leopards and their natural prey is of vital importance to ensure the persistence of leopards in both protected and unprotected areas.

Study limitations

Potential criticisms of our study are that data were collected roughly 2 years apart, and that we investigated factors influencing behaviour only at a single spatial scale. We are aware that community structure could be influenced by weather on a seasonal or annual basis. The Western Cape experienced a relatively dry year (South African Weather Service, 2021) prior to the Cederberg survey and recorded precipitation of 212.9 mm below the annual mean (2017; Data provided by www.meteoblue.com) in the region. A precipitation anomaly of 139.9 mm below the mean (2019; Data provided by www.meteoblue.com) was also noted prior to the survey in the Piketberg region. Nonetheless, Müller et al. (2022b) showed that the leopard population in the Cederberg remained relatively stable during the decade prior to, and including, our study period. Currently, the same quantitative insight does not exist for the Piketberg community, highlighting a need for long-term population monitoring. Ideally, ecological modelling should aim to incorporate different orders of scale (e.g. Pitman et al., 2017), but this is not always practical. We concur that spatial scale must, however, be accounted for in conservation decision-making.