*2.2. Procedure*

A total of 30 camera traps were distributed randomly on the catena to give the best cover of each zone. This included the mud wallows on the catena, the nearby graniteboulder outcrops (inselbergs) and the three closest permanent waterholes. Bushnell (23), Cuddeback (4), Scoutguard (2) and Little Acorn (1) models with similar trigger response times and field of view ranges were used [15]. Cameras were operational in the growing season for a period of two weeks each during September 2015, March 2016 and March– April 2017 in order to compile the basic list of mammals present and to compare the data of similar survey periods spanning three years. All cameras were checked after one week of operation, and memory cards were replaced. The same positions were used for the three survey periods. Two photographs were taken when a camera was triggered, with a 5 s rest interval in-between.

Cameras on the catena were positioned in such a way to be able to include small (>mongoose) to mega sized mammals at heights ranging from approximately 0.5 m to 1.7 m (higher cameras pointing slightly downwards) depending on how open the view of the camera was [15]. Cameras were facing in a southerly direction as far as possible. Visual surveys of the area were used to determine the positioning of cameras based on recommendations of a camera trapping specialist. Game trials, open clearings, areas with longer distance view, termite mounds and areas where obvious animal activity was noted were selected. To ensure that the entire study area was represented, some areas with denser vegetation were also included. One camera was placed several meters from each waterhole at an average height of 2 m, facing down to include the entire waterhole in the view. This was done to prevent animal tampering that can cause cameras to malfunction (viz. [20]). The waterhole cameras were set to take photos at 5 min intervals, and this was combined with taking photos when the camera was triggered by animal movement. All cameras used infrared flashes to take images during the night, but the cameras were too far for the flash to capture animals at the waterhole during the dark part of the night. Only larger animals in the vicinity of the camera were captured by the waterhole cameras at night. There were also additional cameras placed in vicinity of the waterholes but not looking directly at the waterhole.

### *2.3. Data Analyses*

Data of all the cameras that were positioned in the same catenal zone were combined to determine animal presence in that zone. The waterholes, mud wallows and graniteboulder outcrops were treated as three additional zones/areas. For the purpose of this study, a trigger event is described as the image/s recorded of an individual or group of a mammal species triggering the motion sensor of the camera at a specific time. If the species could not be identified due to blurred photos, it was indicated as Unknown. The date and time stamps on the photographs were used to determine separate events. A new event was recorded when there was an absence of the group or individual for at least 30 min, approximately.

The type of feeder and body size of each species present were confirmed by Estes [22] and tabulated. The total number of trigger events (in other words the number of observations) of each species was calculated per zone. The data were graphically presented as a colour gradient in different blocks—white blocks indicate no observation of that species, while the darker the colour, the higher the total number of events/observations (range) that were noted for that species. The three survey periods were indicated on this colour figure for each zone (including mud wallows, waterholes and granite outcrops, separately). The number of events was counted for all mammal species observed at each of the three waterholes during a survey period and that was totalled per species. The maximum number of individuals of a species that occurred together in a group at the waterholes during one event was indicated in a table for that survey period. Other smaller groups and solitary individuals were observed for the majority of the species visiting waterholes, but only the maximum group size was tabulated.

The Shapiro–Wilk test was used to test for statistical normality of the data. Since data were not normally distributed, the Kruskal–Wallis test was used to determine significant differences between mammal presence (indicated with a 1 if present and 0 if absent) in each catenal zone and between different survey periods, respectively. A 5% level of significance was used. The species richness was measured by the number of different species observed in each zone during a survey period, while the Shannon–Weiner Index measured the diversity and evenness. The zones were included separately in the calculations and the sum totals of the number of events per species were included for the catena and the waterholes including the surrounding areas. A two-way ANOVA was performed on these values to determine significant differences in number of events between zones and survey periods. The vegan package in R was used for the analyses [23].
