2.4. Diet and prey selection

The diet of leopard was determined by the analysis of the confirmed leopard scats, after washing them in a laboratory and drying the remains. For each scat, we separated hairs, claws, hoofs and bones from different prey species, and estimated per cent volume of each species. Prey items that were 1% or less of scats were excluded from analysis. Hair samples were identified to species by examining the structure of the medullas, which became visible under the microscope after soaking hairs in xylene for 24 h. Medulla structures were compared to those in a reference collection of hairs from known species. Hairs of leopard were found in some scats, although in trace amounts (≤5%), and these were excluded from analysis because we assumed they were ingested while grooming.

Results from scat analysis were quantified in terms of frequency of occurrence (i.e. percentage of scats containing a particular food item) for comparison purposes [46]. However, because the frequency of occurrence can be misleading given that smaller prey species contribute more to a scat than larger species [45,46], we obtained the per cent biomass consumed (BC). The per cent BC, considered the best approximation of the true diet [46], was calculated using a regression equation derived from feeding trials on captive cougar [47]. Cougar is similar to leopard in body size and niche, thus we assumed they would have a similar prey spectrum and therefore the model would be applicable to leopard [46], and this method has been used by previous studies to estimate leopard diet [45,48–52]. In the regression equation, Y_i = 1.98 + 0.035X_i, X is the live body mass of prey, whereas Y is the mass of prey per collected scat (i). The mean live body mass of all prey species was taken from Francis [53]. For all species less than 30 kg, we used the adult weight, or the mean of adult males or females if both were given. If we could not distinguish hairs between two species (e.g. black-shanked douc and Indochinese lutung), then we took the mean of weights given for those two species. For ungulates greater than 30 kg (i.e. banteng and wild pig) we used half the weight given for adult females, assuming leopard killed both adult females and young. We excluded arthropods and unknown mammal (total three scats) from the biomass calculations because we were unable to accurately estimate BC for these prey items.

We calculated Jacobs' index D [54]: D = (r – p)/(r + p − 2rp), where r is the fraction of prey used, and p is the fraction of prey available. Jacobs’ indices were calculated to investigate prey selection using biomass data to assess which prey species were selected (0 < D ≤ 1) and which were avoided (−1 ≤ D < 0). For each prey species, the D-value depends on which other species are included in the calculation; therefore, we calculated D-values only for the three ungulate species that were consumed by leopard. To determine fraction of prey available, we used ungulate densities estimated by WWF Cambodia personnel. Ungulate densities were estimated using distance-based line transect sampling [55], following the same field protocols and transects of Gray et al. [14]. During the dry season of 2014, a total of 38 transects (2–3 km in length) were surveyed four to eight times within the SWS core zone (approx. 1700 km²). A total of 603 km was walked across all transects, with observers walking transects just after sunrise (start time = 05.00–07.30) and just prior to sunset (start time = 15.30–17.30). Data recorded for ungulates included: species; number of animals (cluster size); distance between the animal or centre of animal group and the observers (using laser rangefinders); compass bearing to the animal or animal group; and compass bearing of the transect line. Data were analysed using the program DISTANCE v. 6.2 [56]. In order to fit detection functions, number of encounters per species were pooled across two adjacent protected areas (SWS and Phnom Prich Wildlife Sanctuary (PPWS)) and densities were calculated only for those ungulates that had the minimum of 40 observations [55] (see Gray et al. [14] for more details). Species data were pooled from 2010, 2011 and 2014 to produce a global detection function. Densities were then stratified by protected area and year to produce ungulate densities for SWS. Biomass available for each ungulate species was obtained by multiplying the density estimates by the adult female weights given by Francis [53].

We calculated leopard diet and D-values based on all confirmed leopard scats that were collected, and then calculated diets and D-values for males and females separately based on the subset of scats that were assigned to sex. For scats assigned to different individuals, we calculated and compared the diet of individuals with ≥5 scats for illustrative purposes only.

We calculated predation impact on ungulates from an equation used previously for carnivores [57,58]: N_prey = (D × DFI × B_prey × n_days × 100)/BM_prey, where N_prey is number of prey individuals consumed by leopard/100 km², D is density of leopard/km², DFI the daily food intake of leopard, B_prey is the per cent BC by leopard for a given prey species, n_days the number of days (i.e. 180 days representing the dry season), and BM_prey the mean live body mass of prey. The DFI for leopard was assumed to be 4.01 kg, based on the average of four different studies that calculated DFI for leopard [52,59–61]. The BM_prey was taken from Francis [53] (see above). For each ungulate species, N_prey was divided by density (individual/100 km²) to determine the per cent of ungulate population consumed by leopard. Predation impact could only be calculated for the 6-month dry season, as this was the only season we collected scats. Scats collected from two consecutive dry seasons were pooled in analysis.