The caracal Caracal caracal is a solitary felid and the largest of Africa's smaller felids (<20 kg), with males weighing up to 15 kg and females up to 12 kg (Skinner and Chimimba 2005). Previous caracal diet studies mostly utilised the invasive method of stomach contents analysis of dead caracal, or alternatively non-invasive scat analysis (Grobler 1981, Stuart 1982, Palmer and Fairall 1988, Avenant and Nel 1997, Avenant and Nel 2002, Braczkowski et al. 2012). Scat analysis can overestimate the importance of smaller prey items, such as rodents and invertebrates, in a predator's diet (Klare et al. 2011); however, data analysis methods have been developed to decrease such biases (Ciucci et al. 1996, Marucco et al. 2008, Klare et al. 2011). Other methods used in dietary investigations include stable isotope and fatty acid analyses (Iverson et al. 2004, Thompson et al. 2005). However, these methods typically cannot differentiate between species–specific prey items for complex multi-prey systems (Hardy et al. 2010).

Global positioning system (GPS) technology has advanced carnivore research substantially (Cagnacci et al. 2010) with increased accuracy of kill site identification (Bacon et al. 2011, Martins et al. 2011), as well as providing valuable insight into other aspects of carnivore ecology including movement (Martins and Harris 2013, Odden et al. 2014), resting (Cristescu et al. 2013) and habitat selection (Cristescu et al. 2014, Fattebert et al. 2015). The use of GPS cluster visitation as a method to estimate carnivore diet has been reported to over-represent larger prey items (Pitman et al. 2012, Tambling et al. 2012, Clark et al. 2014). Svoboda et al. (2013) is one of the few studies wherein GPS radio-collars were used to determine and visit kill sites of a mesocarnivore, the bobcat Lynx rufus. This study, however, only focused on locating white-tailed deer Odocoileus virginianus kill sites. Studies in Africa utilising GPS clusters to determine carnivore diet have been restricted to large carnivores, such as leopards Panthera pardus and lions Panthera leo (Tambling et al. 2010, Martins et al. 2011, Pitman et al. 2012). To our knowledge, mesocarnivore diet composition has not been reported using a combination of GPS cluster visitation and scat analyses outside of North America.

Caracal are reported to be one of the main mesocarnivores responsible for livestock predation in South Africa (Avenant and Du Plessis 2008, Van Niekerk 2010). The need to understand their feeding ecology has been emphasized in various past studies that have highlighted their potential for extensive damage to livestock (Avenant and Du Plessis 2008, Bergman et al. 2013, Du Plessis et al. 2015). Yet, management of caracal is often conducted based on assumptions and traditional knowledge, with little scientific evidence contributing to management decisions (Kerley et al. 2017).

The main objective of this study was to compare caracal diet estimation using scat analysis and GPS cluster visitation. We provide the first documentation on the comparative use of these two methods applied to a mesocarnivore species outside North America and hypothesize that the methods of sampling (scat versus field investigation of GPS location clusters) will influence the outcome of caracal diet estimation. We hypothesize that scat analysis will reveal a greater occurrence and biomass of small prey items than GPS cluster investigations, which will show diet to be dominated by large prey items. Based on the results of the methodological comparison, we formulate suggestions for future studies on mesocarnivore diet.

Figure 1.

A map of South Africa (insert) showing the location of the study area in the Northern Cape. Locations of caracal Caracal caracal scats collected are indicated as a triangle and caracal clusters where prey remains were found using GPS cluster analysis are indicated as stars.

Material and methods

Scat analysis

Caracal scats were collected opportunistically, along predetermined walking transects and at GPS cluster sites from radio-collared caracal. Walking transects were approximately 1 km in distance and were primarily set-up to search for livestock carcasses on farmlands, however these transects were also used to collect scats. Scats that were very old (collectively lost original shape, were dull in colour, porous and easily crumbled upon applied pressure) were not collected. To avoid pseudo-replication only two scats were collected at each cluster site (Bacon et al. 2011). Caracal scats were distinguished from those of other species based on segmentation, shape and size (Walker 1996). Because many felids use scat as a means of territorial marking, only half of each scat was collected (Martins et al. 2011).

Scat samples were autoclaved at 120°C for 20min to allow for complete sterilisation, individually placed in a sorting tray and sorted under a fumehood, removing macroscopic fragments (e.g. bones, insects) before washing the remains of the scat in a sieve (Cristescu et al. 2015b) and drying the clean hair for 24h in a fumehood. Hair samples were soaked in 70% ethanol for 24h to ensure no particles were still attached to the hairs before further analysis. Hairs were then rinsed with distilled water and dried in a fumehood for an additional 24h, or until dry.

Macroscopic and microscopic identification were used to identify prey items in each scat. Wild mammalian prey were grouped into four classes, differentiating based on body size: large mammals (>40kg), medium- to large-sized mammals (10–40kg), medium sized mammals (1–10kg), and small mammals (<1kg). Livestock was included as a distinct prey class due to relevance to human-wildlife interactions and conflict potential. Non-mammalian prey were grouped taxonomically as birds, reptiles, invertebrates, fruit/ seeds, and vegetation, with a separate category for unknown items. Mammalian prey was identified to species level by means of cross-sections of hairs. Cross-sections were made by randomly selecting hairs with a pair of forceps, placing them longitudinally in a 3 mm plastic Pasteur pipette and following Douglas's protocol (1989). A light microscope was used to photograph and examine slides at 20× magnification (where possible 40×). We used LAS Core ver. 4.0 software to measure cross-sections of the hairs for comparison with the reference collections (Rhodes University, Anita Meyer [The Cape Leopard Trust], Keogh (1979, 1983) and personal slides made from hair collected from carcass specimens encountered in the field). Teeth found in scat samples were used to identify rodents to species level (de Graaff 1981) to further validate hair-based identification.

Table 1.

Biomass of prey consumed by caracal estimated from feeding sites (n = 91) visited in Namaqua National Park and surrounding farmlands, Northern Cape, South Africa. Prey items that could not be identified to species level were not included in calculations. Only prey items that contributed >5% to the corrected biomass are included in the table. For a full list of prey items consumed, refer to Supplementary material Appendix 2.

The FO, corrected frequency of occurrence [CFO] (frequency of occurrence per scat) and percentage biomass were calculated. The use of CFO was recommended by Klare et al. (2011), where each scat has a total weighting of 1. If two prey items were present in one scat, each prey item would receive a weighting of 0.5 and less as the number of prey items per scat increases.

While both FO and CFO include rare food items, biomass establishes the importance of a food item in the diet of the target animal (Klare et al. 2011). To estimate the biomass of prey consumed by caracal we used Baker et al. (1993) linear regression equation developed for bobcat to calculate a correction factor for each prey item:

Where y is the weight of prey consumed per scat collected (kg scat^–1) and x is the average body weight of the prey item (kg) (Bacon et al. 2011). This equation is only applied to prey weighing ≤4.5 kg as this is the weight at which a bobcat would ingest the entire prey item. As with bobcats, caracals only feed on parts of larger prey species such as ungulates (Baker et al. 1993, Skinner and Chimimba 2005). To account for this, Baker et al. (1993) used a set correction factor of 27 for larger prey items to account for bobcats only feeding on part of the prey item. A linear regression equation developed for bobcat was used as no data exists for caracal, and bobcats are a felid species comparable both morphologically and ecologically to the caracal (Baker et al. 1993, Skinner and Chimimba 2005, Macdonald et al. 2010).

Predator diet was analysed from GPS cluster sites with confirmed prey remains (n=91) and caracal scats (n=250). Differences in prey species and prey categories between the two sampling methods were tested with Fisher's exact tests, reporting χ² goodness of fit statistics using the programme STATSoft Statistica 11 (Statsoft Inc.).

Results

Between March 2014 and April 2015, 458 caracal GPS cluster sites were visited. Of these, prey remains were located at 91 sites, with a 19.9% success rate of finding prey remains. Cluster visitation yielded a much lower variety of prey items when compared to the 250 scat samples analysed. Only 8 prey items were identified from GPS cluster sites (or consumption sites), whereas 31 prey items were identified in scat samples (Fig. 2, Supplementary material Appendix 1). Rock hyrax was the main prey item recorded from GPS cluster site visitations (39.6%), as well as scat analysis (31.2%). Based on body size classes, small mammals (<1 kg) were not identified at GPS clusters, but 29.1% were found in scats analysed (χ² = 71.45, DF = 1, p < 0.001). Small mammals were the second most frequently occurring prey class as analysed from scat. However, according to GPS cluster analyses, livestock (28.6%) was the second most frequently occurring prey class, after medium mammals (56%) (Supplementary material Appendix 2). There was a significant difference in livestock occurrence between the two sampling methods (χ² = 22.72, DF = 1, p < 0.001).

Sheep Ovis aries was the prey item occurring second most frequently at GPS cluster visits (25.3%), significantly higher than what was identified from scat analysis (2%) [χ² = 21.99, DF = 1, p < 0.001] (Table 1). Medium to large mammals, including duiker, steenbok and klipspringer, showed no significant difference in occurrence in caracal diet between the two sampling methods (p > 0.05).

Figure 2.

Comparison of prey items recorded in caracal Caracal caracal diet using scat analysis and GPS cluster visitation. Reptiles (Squamata) and invertebrates (Coleoptera, Orthoptera, Scorpiones and Solifugae) were grouped due to the low percentage of occurrence.

The total biomass consumed according to GPS cluster visitations was 564.4kg, compared to 2344.9kg from scat analysis. Sheep contributed a substantial proportion of biomass to caracal diet (59.5%) as analysed from GPS cluster sites, but only contributed 5% to the total biomass consumed based on scat analysis (Table 2, Supplementary material Appendix 3). Based on GPS clusters, rock hyrax contributed 17.2% to the total biomass consumed. However, rock hyrax contributed the bulk to the biomass from scat analysis (35.8%). Otomys spp. and Namaqua rock mouse Aethomys namaquensis were absent at GPS clusters but contributed 13.4% to the total biomass consumed as analysed from scats.

From the 250 scats collected and analysed, 89 (35.6%) scats were collected at GPS cluster sites (both feeding and non-feeding sites), 129 (51.6%) opportunistically across the study area and 32 (12.8%) along predetermined transects. Only 17 scats (6.8%) were collected at GPS cluster sites where prey remains were found. Prey classes identified in scats were generally evenly distributed between the manner in which these scats were collected (Fig. 3). From the 89 scats collected at GPS clusters, 36.1% of the prey classes identified were small mammals. The overall patterns of caracal diet remained consistent between the three scat collection methods, with medium mammals and small mammals being the main prey classes consumed by caracals in the study area.

Table 2.

Biomass of prey consumed by caracal estimated from scat (n = 250) collected in Namaqua National Park and surrounding farmlands, Northern Cape, South Africa. Only the relative biomass consumed and only prey items that contributed >5% to the biomass consumed are indicated. For a full list of prey items consumed, refer to Supplementary material Appendix 3.

Discussion

Prey composition identified from GPS cluster visitations and scat analysis differed, with a higher number of prey items identified through scat analysis. GPS cluster visitation findings represented primarily larger- and medium-sized prey items, such as rock hyrax and Lagomorpha, but smaller prey items were not recorded. Past studies found that small mammals, especially rodents, are prominent prey items in caracal diet (Stuart and Hickman 1991, Avenant and Nel 1997, Avenant and Nel 2002, Mellville et al. 2004). Scat analysis in our study further emphasizes the role of small mammals in caracal diet as an important prey item. The use of a corrected biomass consumed provided insight into which prey items are quantitatively important to the predator (Klare et al. 2011). In our study, smaller mammals (≤1–10kg) identified from scat analysis contributed 25.3% to the total biomass consumed, making it the second most important prey class in caracal diet. However, this prey class was completely absent from GPS cluster site analyses. Martins et al. (2011) used GPS cluster visitation and scat analysis to determine leopard diet in the Cederberg Mountains of South Africa. Leopards in the Cederberg are exposed to a limited and small-bodied prey base, consequently resulting in a preference for smaller prey items such as medium-sized ungulates, rock hyrax and even small mammals (Martins et al. 2011). Similar to our Namaqualand study, Martins et al. (2011) found that small mammals were absent from GPS clusters investigated, despite having 20% occurrence in scats.

Figure 3.

Comparison of scat collection methods showing the frequency of occurrence (FO) [%] of all prey classes identified from the 250 caracal Caracal caracal scats analyzed from Namaqua National Park and surrounding farmlands, Northern Cape, South Africa.

We found a low contribution of medium-large wild ungulates to caracal diet based on both scat and cluster site analyses; however, sheep was a prey item that contributed substantially to the total biomass consumed by caracal in the study area. This might be a product of high numbers of sheep in the region (Jansen 2016) and in the hypothetical absence of small-stock, it is possible that caracal diet would include a larger proportion of wild ungulate prey than we documented (Grobler 1981, Braczkowski et al. 2012). According to GPS cluster visitations, sheep contributed 59.5% to the total biomass consumed, compared to only 5% as analysed from scats. If GPS cluster visitation was used as the sole method of dietary analysis, the importance of livestock as a prey item may have been overestimated. Our study only had male caracal collared, whereas caracal scat samples included males and females. Male felids have been generally found to be more predisposed to livestock predation than females (Loveridge et al. 2010). Male felids hold larger territories and young males disperse great distances to establish home ranges, making them more predisposed to contact with livestock (Linnell et al. 2001, Loveridge et al. 2010). In future work, DNA analysis of scats could be used to determine the sex of the individual that deposited each scat (Oyler-Mccance and Leberg 2012). This approach could answer the question as to whether in fact male caracal are mostly involved in livestock consumption in Namaqualand.

Several recent studies on large carnivore diet have used a combination of GPS cluster visitation and scat analysis (Bacon et al. 2011, Morehouse and Boyce 2011, Tambling et al. 2012, Pitman et al. 2013, Cristescu et al. 2015b). To our knowledge, only one study utilising GPS cluster visitation methods to determine the diet of medium-sized felids has been published to date (Svoboda et al. 2013). However, the Svoboda et al. (2013) study focused solely on bobcat predation on white-tailed deer and concluded that rapid visitation of cluster sites by researchers after the bobcat left the area should be a priority, especially to account for smaller prey species, such as white-tailed fawns. The success of finding prey remains when visiting GPS clusters for our study in Namaqualand was slightly higher (19.9%) than what Svoboda et al. (2013) found for bobcat kill clusters in general (17.4%). In Namaqualand we found it to be faster to locate ungulate carcasses at caracal consumption sites than finding rock hyrax or Lagomorpha carcasses, of which only the rumen, intestines and hair tufts were not consumed by caracal. The scavenger community likely influenced our estimation of caracal diet. Scavenger presence at caracal clusters presumably decreased detection of prey remains, or led us to overestimate biomass intake by caracal if the scavengers consumed large parts of a carcass. We suspect that scavenger effects on the reliability of our dataset were not overly extensive, because with the exception of black-backed jackal and some raptors, scavengers in our study area are small-bodied carnivores which caracal would likely displace at a kill.

Due to the sensitive nature of human–wildlife conflict concerning depredation, it is important to report results with caution. Had this study only presented GPS cluster results, it would have indicated that sheep make up the bulk of caracal diet in Namaqualand – an outcome which could lead to increased persecution of caracal. We contend that for medium-sized felids it is important to also include scat analysis if GPS cluster investigations are being used to report predator diet composition. By itself, scat analysis would have likely underestimated the contribution of large prey items to caracal diet. Furthermore, GPS cluster technology can also be used to investigate individual diet/specialization (Elbroch and Wittmer 2013) as well as identify habitat preferences for prey consumption (Cristescu et al. 2019), making it a critical tool in studies on the effects of predators on wild and domestic prey.