Complementing GPS Cluster Analysis with Activity Data for Studies of Leopard (Panthera pardus) Diet

1. What they did

Most leopard diet studies in southern Africa come from savanna and forest, using scat analysis or direct observation. In rugged mountains such as the Cederberg, both are difficult. Researchers can instead look for places where a collared leopard’s Global Positioning System (GPS) locations cluster, then search those spots for prey remains. But it is hard to tell a site where a small animal was eaten from a site where the leopard simply rested. The authors tested a new idea: adding activity data, recorded by a motion sensor in the same collar, to GPS cluster analysis.

• Collaring. The Cape Leopard Trust captured an adult female leopard, F10, on 18 June 2008 and fitted a GPS collar that recorded 8 to 24 locations a day. The collar also carried a dual-axis acceleration sensor that recorded activity every 5 minutes.
• Finding clusters. They mapped the locations and marked clusters of two or more locations within 100 m over at least 4 hours as possible kill sites.
• Searching for kills. At each visited cluster, they searched a 100 m radius thoroughly and collected bones, hair, horns, feet and hooves to identify the prey. A site with no remains after an intensive search was called a non-kill cluster.
• Analysing activity. They compared activity in the hours before and after each cluster began, between kill and non-kill sites and between small kills (under 5 kg) and large kills (over 5 kg). They also looked for repeating rhythms in activity at eight kill clusters and during eight periods when the leopard was moving normally.

This was presented as a pilot study, as only one leopard had enough GPS, activity and field data at the time.

2. What they found

Kills

• Over 254 days, the collar gave 1760 GPS locations (87% fix success) and about 70 000 activity values.
• Of 78 possible kill sites, 54 were visited, on average 171 days (range 22 to 302 days) after the presumed kill. Prey remains were found at 31, a success rate of 57.4%.
• Rock hyraxes (55% of kills) and klipspringers (22.6%) made up most of the diet. Both are daytime animals of rocky, rugged ground. Common duiker and Cape grysbok made up another 9.7%. One Cape porcupine and one Hewitt’s red rock rabbit were also found. All prey weighed under 20 kg.
• By biomass, klipspringers (39.7%) and rock hyraxes (29.3%) contributed most. The leopard spent on average 7.7 hours at hyrax kills and 29 hours at klipspringer kills.

Activity

• Activity generally dropped once a cluster formed.
• Activity at kill and non-kill clusters did not differ significantly.
• Activity at small and large kills was similar before the kill but differed after it, though not quite significantly (P = 0.07). At small kills, activity dropped sharply and stayed low for at least seven hours. At large kills, activity rose markedly within about two hours, presumably from eating.
• During normal movement, most activity followed a 24-hour (daily) rhythm, but this rhythm was mostly missing at kills. Short rhythms of about 3 to 5 hours (periods of 2.7 to 5.3 hours) were more common at kills than during normal activity, significantly so at several frequencies.

3. What they concluded

• GPS clusters alone can show parts of predation such as diet preferences and handling time. Combined with activity data, they can help tell kills from non-kills, especially for larger prey where the leopard stays more than 24 hours.
• At large kills, the leopard’s activity followed a distinct pattern: rest with low activity, broken by feeding bouts of varying length. No distinct feeding pattern was found at small kills.
• Very low or no activity throughout a cluster suggests a resting site, so the activity sensor could help rule out non-kill clusters before fieldwork and reduce field effort.
• Short visits (under about 16 hours) were rarely found to be kills, and could be resting sites or small-prey kills, so the authors could not work out an overall kill rate for all prey sizes.
• The method is laborious: GPS and activity data are recorded at different intervals and must be matched, and irregular GPS schedules complicate analysis.
• The diet results agree with an earlier Cederberg study and with scat analyses.
• Improvements: more frequent GPS fixes, quicker and timely visits to clusters (kills are easier to find soon after), real-time GPS data, and more leopards of each age and sex.

4. How tracking was or was not used

No trailing was reported. GPS clusters were searched on foot within a 100 m radius for prey remains, and the prey species was identified from bones, hair, horns, feet and hooves. Kill sites were confirmed only by finding these remains. The authors note that remains in the Cederberg could be washed away, covered by vegetation or burnt, but that with few scavengers in the area bone fragments would still remain; they report locating kill remains up to 3.5 years after the kill (citing Martins et al. 2011).

5. Tracker or observer certification

Not reported. The paper does not say who searched the clusters or identified the prey remains. Willem Titus is thanked for help with field investigations. The diet results depend on these identifications.

6. Other technologies, skills or innovations

• Vectronic GPS-PLUS collar (Vectronic Aerospace, Berlin), weighing 2% of the leopard’s body weight, with a built-in dual-axis acceleration sensor recording activity every 5 minutes.
• ArcGIS 9.2 to map and inspect location clusters.
• The Lomb-Scargle periodogram, a method from astronomy for finding repeating cycles in unevenly spaced data, used here to pick out feeding rhythms at kills.

7. Implications for conservation (from the paper)

• Combining GPS telemetry with activity sensors offers a useful way to study the feeding of wide-ranging, elusive predators in remote, rugged areas where direct observation and scat collection are difficult.
• Detecting small prey matters to avoid kill-rate estimates that are biased towards larger prey.