LIMPOPO LEOPARD MONITORING REPORT CAMERA-TRAP SURVEY 2016 | DINOKENG LIMPOPO LEOPARD MONITORING PROJECT DINOKENG CAMERA-TRAP SURVEY 2016 Ross Pitman*, Gareth Mann, Gareth WhittinGton-Jones, Lisa Thomas & Guy BaLme * [email protected] (Corresponding author) INTRODUCTION The Limpopo Leopard Monitoring Project aims to provide robust data on leopard population trends in Limpopo Province to inform conservation policy and management. Here we report on a camera-trap survey undertaken in Dinokeng Game Reserve (hereafter ‘Dinokeng). This is the first annual survey for Dinokeng. METHODS We deployed paired camera stations were setup at 36 locations for a total of 48 days. To ensure all individuals within the sampled areas had a probability > 0 of being captured, camera-traps were distributed an average of 2–3 km from one another. To maximize the probability of photographing leopards, camera-traps were placed in high-use areas, such as drainage lines, animal paths, and roads. Camera-traps were mounted on trees or steel poles located 2–4 meters from the focal movement pathway. To reduce false photographic captures, we cleared any vegetation that might obstruct the camera-trap’s field of view. Camera-traps were not moved during the surveys. Camera-trap images were catalogued using camtrapR (Niedballa et al. 2016), within the R Statistical Environment (R Core Team 2015). We identified individuals based on their unique pelage patterns within the pattern recognition software, Wild-ID (Bolger et al. 2012). In addition, all computer-assisted identifications were manually verified. Bayesian spatially-explicit capture-recapture models We followed the capture re-capture analytical methods, and hierarchical model formulation, described by Goldberg et al. (Goldberg et al. 2015) and Royle et al. (Royle et al. 2009). The model relates the observations, yijk, of individual i in trap j during sampling interval k to the latent distribution of activity centers. Observation, yijk, took the value of one for a capture, and zero if not captured, to produce a 2 | Page capture history for all individuals in all traps over all sampling intervals. Multiple detections of the same individual, within the same sampling period, were taken as a single capture. Individuals could be captured on multiple traps during a sampling interval (24 hours). We followed the formulation of the observation process used by Goldberg et al. (Goldberg et al. 2015), Gardner et al. (Gardner et al. 2010), and Russell et al. (Russell et al. 2012). Our spatially-explicit capture-recapture models were implemented within a Bayesian framework using data augmentation (Royle & Young 2008; Goldberg et al. 2015). Data augmentation adds a sufficiently large number of all-zero capture histories to create a dataset of size M individuals (Goldberg et al. 2015). Augmentation was considered large enough when the number of augmented individuals did not truncate the posterior estimates of population size (Goldberg et al. 2015; Proffitt et al. 2015). Data augmentation in this study was set to 400. We chose a uniform prior distribution from 0 to M on population size (Goldberg et al. 2015). Starting values for parameters were: s = 1, q = 0.75, ln(a0) = 0, b = 0, Y = 0, Ysex = proportion of males sampled. We used improper priors (-¥,¥) for a0 and all b parameters, (0, ¥) for s, (0.5, 1) for q, and (0, 1) for Y and Ysex. Models were fit using Markov chain Monte Carlo (MCMC) methods within R, using the SCRbayes package (available at: https://sites.google.com/site/spatialcapturerecapture/scrbayes-r-package). To account for individual, sex-specific effects, we included a sex covariate within all models. Although cubs (< 12 months old) were occasionally captured on the camera-traps, we only included adults and sub-adults within our analyses. To account for heterogeneity in habitat use across the study area, we modelled our density estimate using an existing resource selection function (Pitman et al. in press) as a density covariate (Royle & Chandler 2013; Proffitt et al. 2015). All analyses were run using a statespace of 20 km. Models were run for 50,000 iterations, with a burn-in of 10,000. To reduce autocorrelation, we thinned the MCMC chains by skipping every other iteration, resulting in 12,500 iterations in our posterior sample. We evaluated model goodness of fit using a standard Bayesian P-value approach (Royle et al. 2013). Convergence of the MCMC chains were assessed by examining posterior parameter-wise traceplots and histograms. The mean and 95% credibility intervals, for each model parameter, were then computed from these converged samples (Goldberg et al. 2015). In addition to estimating population density, we assessed the demographic composition of the sampled population. We estimated the age and sex of captured leopards using their relative body dimensions, the presence of a well-developed dewlap, and facial scarring (Balme et al. 2012). We classified leopards into 3 | Page three age classes: juveniles (≤ 2 years), subadults (>2 years; ≤3 years) and adults (>3 years). For adult males, we distinguished between individuals <7 years and ≥7 years. RESULTS Camera-trap surveys: The total area covered by camera-trap stations at Dinokeng amounted to 173.4 km2. The survey ran from the 02 July 2016 to the 18 August 2016, and sampling effort comprised 2,968 camera-trap nights. A total of 81,109 photographs were recorded, of which 62,222 were independent captures (this includes duplicates, ‘blank’ photos, and photos of the research team). A total of 58 different species were recorded (see Appendix 1 for a summary). Unfortunately, leopards were not photographed at any camera stations across Dinokeng, and suggests that no resident leopard population exists on the reserve. Other medium and large carnivores were, however, captured on numerous occasions. Lion were captured on 129 occasions (Fig. 1) across Dinokeng, whilst brown hyaena were captured on 49 occasions (Fig. 2) predominantly within the northern-western section of the reserve. Cheetah were infrequently captured on Dinokeng (n = 3; Fig. 3), but this is in large part a result of our camera-trapping array, which was not designed to photograph cheetah. Caracal and serval were captured fairly frequently on Dinokeng, amounting to 19 (Fig. 4) and 52 (Fig. 5) captures, respectively. Dinokeng Game Reserve Captures 5 10 15 Dinokeng_2016 -25.25 -25.30 -25.35 -25.40 -25.45 -25.50 28.30 28.35 28.40 28.45 28.50 4 | Page Figure 1. Lion capture frequencies recorded at camera-trap stations in Dinokeng during the 2016 survey. Larger circles indicate greater lion activity. Dinokeng Game Reserve Captures 2.5 5.0 7.5 10.0 12.5 Dinokeng_2016 -25.25 -25.30 -25.35 -25.40 -25.45 -25.50 28.30 28.35 28.40 28.45 28.50 Figure 2. Brown hyaena capture frequencies recorded at camera-trap stations in Dinokeng during the 2016 survey. Larger circles indicate greater brown hyaena activity. 5 | Page Dinokeng Game Reserve Captures 1.00 1.25 1.50 1.75 2.00 Dinokeng_2016 -25.25 -25.30 -25.35 -25.40 -25.45 -25.50 28.30 28.35 28.40 28.45 28.50 Figure 3. Cheetah capture frequencies recorded at camera-trap stations in Dinokeng during the 2016 survey. Larger circles indicate greater cheetah activity. Dinokeng Game Reserve Captures 12 3 4 Dinokeng_2016 -25.25 -25.30 -25.35 -25.40 -25.45 -25.50 28.30 28.35 28.40 28.45 28.50 Figure 4. Caracal capture frequencies recorded at camera-trap stations in Dinokeng during the 2016 survey. Larger circles indicate greater caracal activity. 6 | Page Dinokeng Game Reserve Captures 2.5 5.0 7.5 10.0 Dinokeng_2016 -25.25 -25.30 -25.35 -25.40 -25.45 -25.50 28.30 28.35 28.40 28.45 28.50 Figure 5. Serval capture frequencies recorded at camera-trap stations in Dinokeng during the 2016 survey. Larger circles indicate greater serval activity. As the primary leopard prey species, impala and kudu were captured throughout the reserve, and on 1,202 (Fig. 6) and 383 occasions (Fig. 7), respectively. 7 | Page Dinokeng Game Reserve Captures 50 100 150 200 Dinokeng_2016 -25.25 -25.30 -25.35 -25.40 -25.45 -25.50 28.30 28.35 28.40 28.45 28.50 Figure 6. Impala capture frequencies recorded at camera-trap stations in Dinokeng during the 2016 survey. Larger circles indicate greater impala activity. Dinokeng Game Reserve Captures 20 40 60 Dinokeng_2016 -25.25 -25.30 -25.35 -25.40 -25.45 -25.50 28.30 28.35 28.40 28.45 28.50 Figure 7. Kudu capture frequencies recorded at camera-trap stations in Dinokeng during the 2016 survey. Larger circles indicate greater kudu activity. 8 | Page Analyses of activity patterns indicate that lions on Dinokeng are active during nocturnal hours, with peaks at crepuscular periods (Fig. 8)—with a noticeable peak during early morning hours. In contrast, impala avoided moving around during nocturnal hours, and rather opted to remain active during daylight hours (Fig. 9). Caracal and serval appeared active during similar times, which is clearly evident in the large degree of temporal overlap (grey regions; Fig. 10). Lion and brown hyaena similarly exhibited large degrees of temporal overlap, though brown hyaena clearly favoured nocturnal hours, whilst lion showed a peak of activity during post-dawn hours (Fig. 11). Activity of Lion number of records: 129 0.08 0.06 Density 0.04 0.02 0.00 0:00 6:00 12:00 18:00 24:00 Time Figure 8. Lion activity patterns in Dinokeng during the 2016 survey. Higher peaks indicate increased activity, whilst troughs indicate periods of lower activity. 9 | Page Activity overlap: Lion − Impala number of records: 129 / 1202 Lion Dhat1=0.4 Impala 0.10 0.08 0.06 Density 0.04 0.02 0.00 0:00 6:00 12:00 18:00 24:00 Time Figure 9.