Introduction

Small carnivores (carnivora < 16 kg) can play an important role in ecosystem function (Roemer et al. 2009), yet there is a paucity of information on habitat use and basic ecological requirements for many species (Easter et al. 2020; Marneweck et al. 2021). Despite facing similar threats to apex carnivores (e.g. habitat alteration, unsustainable hunting, disease; Marneweck et al. 2021), these smaller carnivore species are often overlooked as a focus of ecological research (Brooke et al. 2014; Hardouin et al. 2021). The lack of research attention on small carnivores creates a disparity between the conservation effort placed on these species when compared to larger, apex predators (Marneweck et al. 2021). This disparity in research focus is exemplified in Africa, where the population trends of most small carnivore species are data deficient (Do Linh San et al. 2013) and ecological data on these species are scarce (de Satgé et al. 2017; Easter et al. 2020). As effective conservation management requires detailed knowledge of species-specific habitat requirements, small carnivore conservation is often hampered by this shortage of available information (du Plessis et al. 2015).

The presence of dominant species (i.e. apex predators) is often a key driver of small carnivore behaviour (Ritchie and Johnson 2009). Competition for resources and the threat of intraguild predation can influence species’ ecology, shaping habitat use, distribution, foraging strategies and, ultimately, survival (Linnell and Strand 2000; Caro and Stoner 2003). Smaller carnivores can display a wide range of spatial, temporal, dietary and behavioural adaptations to facilitate coexistence with dominant predators (Schuette et al. 2013; Ramesh et al. 2017b). However, the intraguild dynamics between small and large carnivores, and the behavioural mechanisms employed to enable co-occurrence, are poorly understood in African carnivore communities (Ramesh et al. 2017b; Easter et al. 2020).

The caracal (Caracal caracal) is a widespread small carnivore that occupies a diverse array of habitats (Avgan et al. 2016). Despite their widespread distribution, there is a lack of research on caracal ecology and their ability to coexist with the large carnivore guild across their geographic range (Avenant and Nel 2002). In Africa, research on caracal ecology has primarily focused on savannah and agricultural habitats in southern Africa (Avenant and Nel 1998; du Plessis et al. 2015; Ramesh et al. 2017a), whilst information on the species’ habitat requirements from other areas has been considerably overlooked. In this study, we investigate caracal habitat use and temporal activity in a miombo woodland habitat, providing the first information on caracal site use from this habitat, and from Malawi, whilst assessing caracal spatiotemporal responses to the presence of large carnivores.

Methods

Kasungu National Park (KNP) is a 2316 km2 protected area in the central region of Malawi (Fig. 1). The primary habitat in KNP is miombo woodland, consisting of Brachystegia and Julbernardia spp., interspersed with seasonally wet grasslands (Bhima et al. 2003). As a result of unsustainable poaching rates, KNP has experienced a widespread decline in large mammal and carnivore populations (Munthali and Mkanda 2002), resulting in a large carnivore guild that consists of only leopard (Panthera pardus) and spotted hyaena (Crocuta crocuta) (Davis et al. 2021a).

Fig. 1
figure 1

(a) Camera trap locations for surveys conducted in 2017 and 2018 in Kasungu National Park, Malawi. Inset maps show (b) the area covered within Kasungu National Park; and (c) the location of Malawi within sub-Saharan Africa

Full size image

Between 2016 and 2018, large carnivore camera trap surveys, designed to estimate leopard and spotted hyaena density, were conducted in KNP, further details of which can be found in Davis et al. (2021a). We used bycatch data from these annual camera trap surveys to investigate caracal habitat use and temporal activity in KNP. Due to the low sample size for caracal captures in 2016, because of limited camera traps, we only used the 2017 (n = 50 camera traps) and 2018 (n = 25 camera traps) datasets for this study. Most cameras were placed on roads (n = 69), with a limited number placed on game trails (n = 6). We employed an occupancy modelling framework to assess caracal habitat use and developed a binary matrix of detection/non-detection capture histories for caracals at each camera trap station in KNP, where ‘1’ indicated a caracal detection during a sampling occasion and “0” indicated no detection (MacKenzie et al. 2017). Sampling occasions were pooled into fifteen-day intervals to account for a limited sample size and reduce non-detections (MacKenzie and Royle 2005). Due to the small sample size for cameras on game trails, and absence of caracal captures on these cameras resulting in occupancy models failing to converge, we removed these cameras from the dataset.

Single-season, single-species models were produced using the package ‘unmarked’ (v1.1; Fiske and Chandler 2011) in R v4.1.1 (R Core Team 2021). We followed a two-step modelling process, whereby we first identified the detection covariates that best explained heterogeneity in detection probability (p). To test the effect of detection covariates (see Table 1) on caracal detection probability, we varied p using all possible combinations of detection covariates and the most parametrised covariate model for site use (Ψ). We limited candidate models to a maximum of three detection covariates to avoid overparameterisation. We identified the detection model with the lowest AIC value and retained these detection covariates for testing the significance of Ψ covariates. Secondly, using only the top-ranked detection covariates identified in the previous stage, we modelled caracal occupancy as a function of site-specific covariates: (i) percentage grass cover; (ii) relative abundance index (RAI) of leopard and (iii) spotted hyaena; and (iv) distance to water (km) (see Table 1). The use of the RAI covariate for large carnivores was chosen over a co-occupancy modelling approach based on the widespread distribution of leopard and spotted hyaena in KNP, with both species present at ~ 90% of camera trapped sites (Davis et al. 2021b). As there was limited heterogeneity in site use, we reasoned that the intensity of site use, provided by RAI, was a more appropriate metric. Prior to analyses, continuous site use covariates were standardised to a mean of zero and a standard deviation of one. Covariates were tested for collinearity using Pearson’s correlation test, excluding covariates from the same model if correlated at r > 0.7 (Dormann et al. 2013).

Table 1 Covariates used in the occupancy modelling framework that we predicted would influence caracal detection probability (p) and site use (Ψ). A description of how these data were sourced and justification for their inclusion in occupancy models are presented
Full size table

We used a model selection approach to determine the importance of Ψ covariates by comparison of AIC values. Only models with substantial empirical support (ΔAIC < 2) were retained in the final model set (Burnham and Anderson 2002). For the final model set, we applied model averaging, using the ‘MuMIn’ package (Barton 2020), to obtain average β-coefficient estimates and determine the importance and direction of site use covariates. A covariate was considered an important predictor of site use if the 85% confidence interval values of the β-coefficient estimate did not overlap zero, as recommended in Arnold (2010). Model fit and overdispersion (where a value of ĉ > 1.1 is considered overdispersed) were evaluated using a goodness-of-fit test with Pearson’s chi-square statistic and 10,000 parametric bootstraps (MacKenzie and Bailey 2004).

To understand the activity patterns of caracal in KNP, we used the time stamps from photographic captures, converted to radian units and estimated daily activity levels using the Kernel circular density function (Ridout and Linkie 2009; Rowcliffe et al. 2014). We used the ‘overlapEst’ function in the package ‘overlap’ v.0.3.2 (Meredith and Ridout 2014) to quantify the degree of overlap between caracal and large carnivores (leopard and spotted hyaena). Following Meredith and Ridout (2016), we used the Δ4 estimator for the overlap coefficient as all sample sizes were > 75 observations. We used the function ‘compareCkern’ in the package ‘activity’ v.1.3. (Rowcliffe 2019) and 10,000 bootstrap samples to test for significant differences between the activity curves of caracal and large carnivores.

Results

Over 4424 trap nights we recorded 113 independent caracal captures on 28 camera stations (41% of sites surveyed). Using 15-day sampling occasions resulted in a combined total of 315 occasions at 69 stations (average: 4.57 occasions per station), with caracal recorded on 61 sampling occasions. We estimated a mean detection probability of 0.41 (SE = 0.06) and a model-averaged probability of site use of 0.49 (SE = 0.16).

After testing each detection covariate, and all combinations thereof, the leopard RAI detection covariate best explained caracal detection probability and was retained for modelling site use covariates (see Table S1; Supplementary File 1). Caracal detectability was lower in areas of higher leopard relative abundance (βleopard =  − 0.98 ± 0.27). Model selection resulted in three models with strong support (ΔAIC < 2; Table 2). Large carnivore RAI had contrasting effects on caracal site use, spotted hyaena RAI had a negative effect (βhyaena =  − 0.97 ± 0.56; 85% CI =  − 1.77 to − 0.17: Fig. 2), whilst leopard RAI had a positive effect (βleopard = 1.01 ± 0.60; 85% CI = 0.15–1.88). Caracal site use increased with higher percentage grass cover (βgrass = 0.60 ± 0.35; 85% CI = 0.09–1.10) and at sites further from water (βwater = 0.75 ± 0.42; 85% CI = 0.14–1.36). A goodness-of-fit test on the global model did not indicate evidence of overdispersion (ĉ = 0.67) or lack of fit (p = 0.81). The c-hat value indicated slight under dispersion, but no adjustments were made based on this result as typically overdispersion is considered more problematic (MacKenzie et al. 2017).

Table 2 Model selection results for caracal (Caracal caracal) site use in Kasungu National Park, Malawi. Only the ten highest ranking models are shown, plus the null model
Full size table
Fig. 2
figure 2

Model-averaged β-coefficient estimates, with 85% confidence limits, of covariates explaining caracal site use (Ψ). Confidence limits that do not cross zero (dotted line) represent a significant effect on caracal site use. Caracal silhouette: Margot Michaud, PhyloPic

Full size image

Overall activity (estimated proportion of time spent active over the daily cycle) was 0.55 (± 0.06, 95% CI = 0.42–0.64) for caracal. Caracal exhibited high temporal overlap with both leopard (Δ4 = 0.88, 95% CI = 0.82–0.97: Fig. 3) and spotted hyaena (Δ4 = 0.81, 95% CI = 0.76–0.90). However, caracal use of the diel cycle was statistically different when compared with spotted hyaena (p = 0.02), with caracal being detected more often than spotted hyaena during the day.

Fig. 3
figure 3

Overlap in activity patterns from camera trap data between (a) caracal and spotted hyaena and (b) caracal and leopard. Number of photographic captures, coefficient of overlap (Δ), 95% confidence intervals and p-values from a test of probability that two sets of circular observations come from the same distribution are presented

Full size image

Discussion

Although important members of the carnivore guild, caracal remain understudied across their range, and little is known about their ecology outside of savannah habitats. Miombo woodlands have been identified as an important habitat type for small carnivores and information on habitat selection from these areas is required to inform conservation management (Hardouin et al. 2021). Our findings indicate that caracal habitat use increases at sites with higher grass cover and those further away from permanent water. Caracal showed spatial and temporal partitioning with spotted hyaena in KNP, with areas of spotted hyaena abundance negatively associated with caracal site use. In contrast, we did not find evidence for spatial or temporal partitioning with leopard across the two niche axes investigated.

Caracal site use increased in areas with higher grass cover, with this positive association likely a result of the caracal’s ambush technique for catching prey (Avenant and Nel 2002). Previous studies have indicated that caracal are habitat generalists (Avenant and Nel 2002), although Mwampeta et al. (2020) found that caracal preferred wooded grassland areas in Serengeti National Park, Tanzania. Sites with higher grass cover in KNP could provide caracal with increased hunting opportunities for important prey, such as rodents and hares (Avenant and Nel 1998; Jansen et al. 2019), whilst also offering refuge from larger carnivores and dense cover for resting sites. In addition, sites further away from permanent water were more likely to be occupied by caracal. Our findings support those of Ramesh et al. (2017a) and Mwampeta et al. (2020) in that caracal habitat use is not reliant on areas close to permanent water. Caracal can often persist in arid conditions and utilising sites further away from permanent water could be a strategy to reduce competition with other medium-sized carnivores, such as serval (Leptailurus serval), that are more reliant on permanent water and compete for similar prey (Mwampeta et al. 2020). Our results suggest that caracal display habitat-specific preferences in miombo woodlands and may exhibit more specialised habitat requirements in areas where apex carnivores are present.

Sites with a higher abundance of spotted hyaena had a negative impact on caracal site use, whilst caracal and spotted hyaena use of the diel cycle was significantly different. Small carnivores generally avoid interactions with dominant carnivore species, as the latter exert direct competitive effects on subordinate competitors (Caro and Stoner 2003). Therefore, utilisation of areas with lower spotted hyaena abundance, sites with higher grass cover, and differences in temporal activity limit the likelihood of costly encounters. Theory on top-down ecosystem regulation and suppressive effects of large carnivores indicates that body size has a strong influence on the intensity of competitive effects (Palomares and Caro 1999; Donadio and Buskirk 2006). As such, it would be expected that the presence of leopard has a greater effect on caracal site use, as the two species are closer in body size and share a higher degree of dietary overlap (Müller et al. 2022). However, recent studies have indicated that the network of intraguild competition and top-down mediated effects are more extensive than previously thought (Prugh and Sivy 2020; Curveira-Santos et al. 2022). Intraguild competition increases under resource limitation (Palomares and Caro 1999), and with the depletion of natural prey (Munthali and Mkanda 2002) and the absence of a resident lion population in KNP (Davis et al. 2021a), spotted hyaena may exert greater top-down influences on sympatric carnivores. To our knowledge, this study is the first to explicitly analyse the intraguild dynamics between caracal and spotted hyaena. Future studies should investigate the impact of spotted hyaena on small carnivore behaviour, as the species’ role in driving small carnivore community dynamics may be overlooked (but see Ramesh et al. 2017b).

The lack of caracal spatiotemporal responses to leopard presence is surprising, as the two species can display high dietary overlap (Müller et al. 2022) and leopard are known to kill caracal when encountered (Curveira-Santos et al. 2022). Given the potential for dietary overlap (Caro and Stoner 2003; Müller et al. 2022), caracal and leopard may be selecting similar habitat characteristics for preferred prey, and this could explain the positive relationship between caracal site use and leopard abundance. Previous studies have indicated that whilst caracal and leopard potentially compete for prey, they can exhibit a high level of spatial and temporal overlap and facilitate coexistence using fine-scale avoidance strategies (Müller et al. 2022). Our dataset was insufficient for modelling fine-scale interactions between caracal and apex carnivores, so it is not known whether caracal use similar behavioural mechanisms to enable co-occurrence with leopard here. However, as caracal detection probability in our study was significantly lower at sites with higher leopard abundance, this could suggest behavioural adaptations to minimise the likelihood of interaction. For example, as leopards frequently use roads for territorial patrols (Braczkowski et al. 2016), caracal could be reducing their use of roads and using vegetation cover to move through areas of higher leopard activity. In addition, leopard in KNP also display spatiotemporal avoidance of spotted hyaena (Davis et al. 2021b) and, by avoiding the same apex predator, this could increase the spatiotemporal overlap of leopard and caracal.

Our study has provided novel information on caracal habitat use in a miombo woodland and is the first study on caracal ecology from Malawi. Caracal have previously been reported in Malawi from Majete Wildlife Reserve (Reece et al. 2021) and Vwaza Marsh Wildlife Reserve but the species’ distribution and population status across the country is poorly understood and requires further research. For example, many of Malawi’s protected areas have experienced a decline in large carnivore populations, with lion often extirpated (Mésochina et al. 2010) and other large carnivores found at low densities (Davis et al. 2021a). The decline of apex predators in these protected areas could aid competitive release from the regulating effect of large carnivores, so status assessments of small carnivores in these areas would be beneficial for conservation management. Finally, basic ecological information and knowledge of intraguild interactions is still lacking for many small carnivore species, both in Malawi and across Africa (Ramesh et al. 2017b; Easter et al. 2020). This paucity of ecological information limits small carnivore conservation and management and we encourage further reporting of these data where possible.