Cheetahs (Acinonyx jubatus) running the gauntlet: an evaluation of translocations into free-range environments in Namibia

Subjects and success definition

Free-hold land managers trapped study cheetahs on Namibia’s commercial farmlands between May 2008 and December 2012 (Fig. 1). They reported animals to the state wildlife department and requested translocation as an alternative to lethal control. In the present cases, government were unable to attend to trapped individuals at short notice and the animals were moved into a registered captive facility (Fig. 1) pending further decision on their management (i.e., permanent captivity vs. release). Contrary to a relocation programme in South Africa (Buk & Marnewick, 2010), Namibian managers did not receive financial rewards for translocating cheetahs instead of killing them.

Based on their specific background and the supporting evidence available, we classified cheetahs as livestock raiders (observed at kills and captured within 24 h of the incident), indiscriminate captures (those suspected of killing expensive game species and cases without evidence of livestock predation involvement), or orphans (trapped at pre-dispersal age without mothers) (Table 1).

We define success to include survival for at least one year, no homing to the original capture site or captive facility, and minimal livestock conflict (≤5 livestock per year) that was considered agreeable for compensation. However, landowners received compensation for any verifiable damage from translocated cheetahs. Reproduction—the ultimate indication of biological success—was not a condition for success, but it remained a key criterion in success evaluations. Although generally desirable (e.g., Boast, Good & Klein, 2015), site fidelity was also not a prerequisite for translocation success because all subjects were released into environments permitting free choice of movement.

Management of cheetahs

Supplemental Information 1 provides a detailed description of captivity conditions, immobilisation protocols, and transportation. When we obtained immediate permission for translocation, cheetahs were only admitted to the captive facility to carry out medical examination and for fitting of tracking technology, followed by transfer to their release site. We minimised captive times as much as was practicable. Prolonged captivity (Table 1) resulted from delays in sourcing of recipient sites, rearing of orphans, presence of juvenile offspring, and issuing of permits. We kept study cheetahs with minimal human contact to avoid loss of natural fear. We assessed the degree of habituation to human presence and vehicles (Table 1) using discrete behavioural responses as classification criteria (Supplemental Information 2).

Licensed veterinary personnel (registered with the Namibian Veterinary Council) immobilised cheetahs to enable health assessments, fitting of tracking technology, recording of body measurements and collection of biological samples (Supplemental Information 1). We marked cheetahs with ID-microchips (subcutaneously) for permanent identification.

We define translocation distance as the linear distance from source site to release locality. We preferred long distances (>200 km) to prevent homing (Linnell et al., 1997; Marker, 2002), but distances were influenced by availability of recipient areas. We released cheetahs using hard releases, i.e., directly from the transport crate after being placed at a permanent water source for several hours, or through soft releases, i.e., following variable periods of acclimation in holding pens (Table 1) that measured 1–500 ha.

Recipient areas and release considerations

Recipient areas (Supplemental Information 3) were privately owned nature reserves that promoted large carnivore population recovery. Translocated cheetahs contributed to ecosystem restoration efforts as well as non-consumptive wildlife tourism. Formal feasibility studies were not available and we selected release sites in correspondence with reserve management and the national wildlife authorities (B Beytell, pers. comm., 2008–2010). The reserves provided permanent water, adequate natural prey compositions, and a known local predator guild (Supplemental Information 3). All sites were located within extant species range, with estimated low densities of <2.0 cheetahs/100 km² at the time that releases occurred (Hanssen & Stander, 2004; Stein et al., 2012).

None of the recipient areas had predator-proof fencing (Supplemental Information 3). Therefore, we awarded preference to reserves providing connectivity with large government PAs (Fig. 1). Because we could not ascertain site fidelity for releases into unconfined areas (Purchase, Vhurumuku & Purchase, 2006; Marker, 2002), we also employed a simple data sharing system to inform land managers of the movements of released cheetahs (Supplemental Information 1) and to enable recapture of individuals if necessary. We released cheetahs with a known livestock depredation background only into the largest reserve (i.e., NamibRand) with highest PA connectivity. Recipient reserves actively supported monitoring efforts and provided important logistic support, particularly in terms of cheetah husbandry in soft release enclosures. Reserves also informed surrounding land managers of translocation events and facilitated conflict mitigation through compensation agreements.

We implemented releases as staggered events to allow for sufficient settling and acclimation of previously translocated cheetahs. We maintained an inter-release interval of >4 months for subsequent releases into the same reserve. We released most cheetahs into areas where they experienced novel competition with spotted hyaena (Crocuta crocuta) but none of the recipient areas had free-ranging lions (Panthera leo). There was no supportive post-release management other than the recapture of one individual and the re-release of one accidentally trapped subject.

Monitoring

Except for sub-adults released alongside their mothers and one female tagged with an ID collar as part of a group release, we fitted all subjects with tracking devices to enable intensive post-release monitoring. External transmitters included Very High Frequency (VHF) radio-collars as well as dual VHF-GPS (Global Positioning System) satellite transmission collars (Table 1). Battery life times were ∼36 months for VHF units and 12–30 months for GPS devices. Transmitter weight was <1.5% of an individual’s body mass. We fitted collars so that they allowed for expected neck growth of pre-prime individuals.

We collected positional data with standard VHF telemetry methods (Amlaner Jr & Macdonald, 1980; Kenward, 2001) and via remote GPS satellite telemetry. During direct observations, we took care not to disturb or displace animals. We attempted to locate VHF-tagged cheetahs at least once per week and telemetry efforts generally concentrated on these subjects. GPS units recorded and relayed 1–6 daily locations. Consequently, daily detection probability was <1.0 for VHF-tagged cheetahs and ∼1.0 for subjects with active GPS trackers.

We adjusted GPS sampling regimes to enable investigation of suspected kill sites and to assess conflict involvement of translocated cheetahs. Data sharing with land managers outside of PAs was an important monitoring element (protocol detail in Supplemental Information 1). If we could not locate a cheetah for several weeks, or GPS units failed early, we employed aerial telemetry from a Cessna 182 aircraft to ascertain the individual’s fate. When direct observations were not possible, we calculated locations with software LOCATE II (Nams, 1990) using triangulation of the strongest VHF signal bearings. Camera trap records and observation reports supplemented tracking information.

The Ministry of Environment and Tourism endorsed all research activities (permit numbers: 1254/2008; 1354/2009; 1459/2010; 1459/2011; 1782/2013; 1888/2014) and an ethics approval was obtained from Manchester Metropolitan University (clearance: RD1-06979434-20130923).

Data analysis

We analysed data with software R v.3.1.0. (R Core Team, 2014) and JMP v.11.0 (SAS Institute, 2013). Spatial analyses were computed with ArcGIS v.10.1 (ESRI, 2013) and QGIS v.2.6-2.8 (QGIS Development Team, 2014). After removing outliers and low-quality GPS fixes, we standardised telemetry data to one location per day (closest to 12:00 GMT) (Mizutani & Jewell, 1998). Although this selection violates assumptions of auto-correlation between subsequent positions (Rooney, Wolfe & Kayden, 1998; Swihart & Slade, 1985), it reduces auto-correlation significantly (De Solla, Bonduriansky & Brooks, 1999) and provides a useful approach to enable comparisons between VHF- and GPS-tagged individuals and between GPS-collared individuals with variable sampling regimes.

To determine the duration of exploratory movements, we calculated 100% Minimum Convex Polygon (MCP) values for 10-day increments starting from the day of release. This 10-day period was progressively shifted by one day for the first year to assess when range values reached stable asymptotes (Weise et al., 2015). We confirmed asymptote stability by determining the date when percentage increase in total progressive 100% MCP range values was no longer significant (Gautestad & Mysterud, 1995). For animals with sufficient sample sizes, we calculated home ranges for periods excluding initial orientation and explorations. We measured home ranges as individually smoothed movement-based 50% and 95% kernel density estimations (following AUC-methods in Cumming & Cornélis, 2012) and 50% and 100% MCPs (Mohr, 1947) using the Spatial Analyst Movement Ecology Tools Extension v.10.2.2 (Wall, 2014).

We used percentage overlap of daily locations with the recipient area (i.e., release reserve) and other PAs (private and/or public conservation areas) as a measure of site fidelity for the first 12 months. We assessed homing behaviour using package Circular in R (Lund & Agostinelli, 2013) by calculating bearing angles and distances between an individual’s last known location and release location relative to their capture site. We adjusted bearing angles to set an individual’s ‘true home’ (i.e., original capture site) direction to 0°. We normalised all distances to a scale from zero to one, representing the distance between a cheetah’s capture and release sites. Cheetahs successfully homed if they moved the entire translocation distance towards the capture location, within 22.5°on either side of true home (Fies, Martin & Blank, 1987), or returned to the captive facility.

We assessed cheetah survival by calculating Kaplan–Meier survivorship estimates with the product limit estimator (Kaplan & Meier, 1958) using a staggered entry design (Pollock et al., 1989). This method allows for inclusion of animals entering the study at irregular intervals (Marnewick et al., 2009) and does not discriminate against individuals with unknown fate when collars fail or deplete. We report cheetah prey from direct, opportunistic observations of hunting events (with binoculars) and from carcasses investigated after detection of GPS clusters that indicated potential kills (Knopff et al., 2009). Therefore, prey results are limited to diurnal hunts and they exclude small prey species with low detectability.

Potential recipient areas in Namibia

Considering cheetah ecology, the outcomes of this study, and those of other published cheetah translocations (Table 2), we modified the ArcGIS-based geo-processing tool CaTSuiT (Lemeris Jr, 2013) to determine cheetah recipient area suitability across conservation areas in Namibia—including national parks (137,451.5 km²), private reserves (4,883.9 km²), and gazetted communal conservancies (160,245.5 km²) (Supplemental Information 4). We defined recipient area suitability to include five conditions (Table 2) at equal parameter weighting. Hence, if an area did not meet all of these conditions, it was considered unsuitable. The rationale for this conservative approach was to maximise chances of post-release survival and reproduction, whilst minimising the risk of conflict and persecution (Table 2).

We were not able to include natural prey as a critical element of area suitability (Hayward et al., 2006; Hayward, O’Brien & Kerley, 2007; Lindsey et al., 2011; Winterbach et al., 2015) because standardised wildlife information was not available for most designated PAs. In addition, Namibia’s PA boundaries rarely obstruct wildlife movements effectively and, therefore, local prey densities may fluctuate significantly and according to seasons. Sufficient prey availability needs to be determined case-specifically prior to any future translocations.

Except for Namibia’s coastline, we considered all habitats as potentially suitable because cheetahs used to occur widespread across the country (Gaerdes, 1974; Marker-Kraus et al., 1996). Furthermore, cheetahs inhabit, and readily adapt to, a wide variety of biomes (Bissett & Bernard, 2007; Caro, 1994; Eaton, 1974; Marnewick et al., 2009; Purchase, Vhurumuku & Purchase, 2006; Skinner & Smithers, 1990; Welch et al., 2015), including arid and hyper-arid ecosystems (Mills, 1984; Rostro-García, Kamler & Hunter, 2015; Sunquist & Sunquist, 2002).

We assessed recipient area suitability in a step-wise elimination process—the order of parameter inclusion had no influence on the final results. We computed two different iterations of the model. First, we included all suitability criteria as outlined in Table 2. Secondly, we excluded site fidelity as a critical suitability parameter, reflecting our original study design.

Sample characteristics

We investigated the outcomes of 15 cheetah translocations, comprising 23 adults (12 males, 11 females) with 10 dependent offspring, for a total of 7,725 monitoring days (adults only) between May 2008 and October 2014. Translocations involved three cases with livestock depredation backgrounds, 15 indiscriminately captured cheetahs, and five individuals released after long-term captive rehabilitation. Prior to release at an average distance of 419.6 km ± 216.1 km SD (range: 71–816 km), we held cheetahs for 1–1,184 days (350.6 days ± 439.0 days SD, n = 23) (Table 1), corresponding with availability of recipient areas, rearing of orphans, presence of offspring, and permit acquisition.

We employed hard releases (n = 13) and soft releases (n = 10). For soft releases, the median acclimation period at recipient sites was 18.25 weeks (range: 1–38 weeks, n = 10). We released cheetahs solitarily (n = 5), as mothers with dependent sub-adults (n = 4, +10 offspring), as sibling groups (n = 7, three events), and as groups artificially bonded during captivity (n = 7, three events) (Table 1). We tracked cheetahs with combined GPS satellite—VHF transmitters (n = 12), VHF-only transmitters (n = 10), while Aju02 had an ID collar as part of a group release. We fitted the GPS unit of Aju40 to Aju41 at a later stage. Based on their habituation to humans (Supplemental Information 2), we classed cheetahs as wild (n = 15), semi-habituated (n = 4), or fully habituated (n = 5) (not tame) at the time of release. All sub-adults released alongside their mothers retained wild characteristics. Except for those individuals released into central Namibia (five adults, three offspring), we released cheetahs (18 adults, seven offspring) into areas where they experienced novel competition with spotted hyaenas.

Survival

Three GPS units failed during the first year. Of the remaining 20 collared adults, 10 died within 12 months of release: eight during the first 71 days. For all collared cheetahs (n = 23), the annual Kaplan–Meier survivorship estimate was 0.57 (95% CI [0.35–0.76]) (Fig. 2). Males and females survived equally well (K-M = 0.56, 95% CI [0.26–0.82], n = 12; K-M = 0.53, 95% CI [0.23–0.81], n = 11 respectively) (Z = − 0.1836, P = 0.8572).

Despite acclimation to recipient areas, soft released cheetahs appeared to have lower survivorship in year one (K-M = 0.40, 95% CI [0.14–0.73], n = 10) than hard released individuals (K-M = 0.67, 95% CI [0.37–0.88], n = 13) (Fig. 3) but not significantly so (Z = 1.31, P = 0.1902). Fully habituated cheetahs (n = 5) had the lowest survivorship (Fig. 3). None of them survived for longer than 205 days. Private landowners (who had not received monitoring data yet) shot four outside recipient areas and a spotted hyaena killed the fifth–giving a median survival period of 2.4 months (range: 0.6–6.8 months, n = 5). Landowners who killed habituated cheetahs reportedly shot them because of their lack of fear of humans, rather than for reasons of conflict (Table 3). The association between first year survival and degree of habituation was significant (χ² = 8.63, df = 2, P = 0.0134, n = 20).

In addition to the early deaths of fully habituated individuals, two spotted hyaenas killed male Aju19 13 days after release. Female Aju07 was re-captured 11 days after release due to apathy. She had become emaciated and dehydrated and succumbed to acute renal failure (I Baines, pers. comm., 2010) >12 months later. After leaving the recipient area, female Aju59 died 19 days post-release due to severe injuries sustained during accidental capture in a gin trap set for control of black-backed jackals (Canis mesomelas). Female Aju18 appeared to have died from shock/exhaustion 14 days post-release—a post-mortem delivered no definitive results. She hunted successfully on the day of her release and found at least two permanent water holes on the recipient reserve. Male Aju03 was found with a broken pelvis/spine near a rocky outcrop after almost one year, but the carcass showed no obvious signs of a hyaena attack.

Following three additional mortalities in the second year (Aju01/38/58 in Table 3) and ‘censoring’ of four cases with depleted tracking collars, the progressive Kaplan–Meier survivorship estimate for adults decreased to 0.40 (95% CI [0.21–0.62]) (Fig. 2). Assessed independently across years, adult survivorship in years one (see above) and two (K-M = 0.70, 95% CI [0.35–0.92], n = 10) did not differ significantly (Z = − 0.703, P = 0.4839). However, after the first three months with highest mortality, cheetah survivorship improved and remained ≥0.80 (Fig. 4), suggesting that the initial post-release orientation/exploration period is particularly important in terms of survival. Excluding cases with unknown outcomes, cheetahs surviving the first 90 days (n = 12) had an 83% chance of surviving one year, and a 38% chance of surviving for two.

Overall, human persecution (n = 7) and spotted hyaena attacks (n = 4) had the highest impact on cheetah survival (Table 3). Eight adults died outside of PAs whereas six died on conservation areas. At the end of this study, the mean known survival period of adults was 12.1 months ± 11.1 months SD (range: 0.4–33.0 months, n = 20). The survival periods recorded for females (10.5 months ±10.2 months SD, range: 0.4–28.0 months, n = 10) and males (13.7 months ± 12.3 months SD, range: 0.4–33.0 months, n = 10) did not significantly differ (Wilcoxon-Mann–Whitney U-Test: U = 58.5, P = 0.5443). There was also no significant difference in mean survival periods of hard released cheetahs (14.8 months ± 11.7 months SD, range: 0.4–33.0 months, n = 11) and soft released ones (8.7 months ± 10.1 months SD, range: 0.6–28.0 months, n = 9) (U = 85.5, P = 0.5173).

Reproduction

Of the six females that survived >90 days, four produced litters after settling into permanent ranges. Two females (Aju56/58) successfully reared offspring released with them (n = 5) and reproduced again (eight cubs). Females Aju02/29 produced another six cubs (n = 6) (Table 3). Median litter size after emergence from lairs was 3.5 (range: 2–5, n = 4). Considering a gestation period of ∼94 days for African cheetahs (Brown et al., 1996) and emergence of cubs from the lair at eight weeks of age (Laurenson, 1994), Aju02 conceived as soon as three months post-release. We observed new litters of Aju29/56/58 13–14 months post-release, when cubs appeared to be 2–4 months old (according to phenotypic characteristics described in Eaton, 1969). The known median survival for newborn cubs was 8.0 months (range: 1.0–14.0 months, n = 14) and at least 64% reached an age of nine months.

In addition, we observed two males (Aju17/38) during courtship behaviour with wild females, but mating events could not be confirmed.

Movements

Explorations and settling behaviour

All translocated cheetahs displayed exploratory movements that extended beyond the boundaries of target recipient reserves and connected PAs (Fig. 5). According to incremental 10-day 100% MCP area progressions, the durations and spatial scales of exploratory movements were highly variable but distinct peaks occurred during the first 6.5 months (Fig. 6). Soft release did not prevent large scale explorations (Fig. 6) which were also not site-specific (Fig. 5).

Measured as the total 100% MCP area utilised, at least five cheetahs roamed >2,000 km² during the first three months after release, and covered >4,000 km² within six months (Table 4). Hard released male Aju26 roamed across 19,743 km² within 112 days (until collar failure) whilst the movements of soft released male Aju30 encompassed 9,048 km² in 180 days. Extensive movements were also not sex-specific. Together with her three sub-adult offspring, Aju56 roamed across 13,596 km² within six months (Table 4). For those cheetahs that established/resumed ranges, settling occurred 13–190 days post-release (median = 93 days) (Table 4), whereas Aju26/30/41/42-44 showed no tendency to settle before collars failed or subjects died (Fig. 6).

Table 4:

Summary movement statistics for cheetahs translocated into free-range environments.

ID	Linear distance (km) moved in first 3 months (no. of fixes)	100% MCPa area (km2) covered		Duration of post- release exploratory movements (days)	Estimated home range size in km2			Percent overlap with recipient reserve (all PAs)	Distance (km) of centroid to release site	Distance (km) of last known location to release site
		3 months (no. of fixes)	6 months (no. of fixes)		No. of fixes	100% MCPa (50%)	95% M-KDEb (50%)
Aju01	Insufficient data	820.2 (54)	1,316.0 (103)	93	92	1,509.0 (284.6)	1,196.3 (108.6)	39 (54)	13.5	24.2
Aju02	Insufficient data	820.2 (54)	1,316.0 (103)	93	92	1,509.0 (284.6)	1,196.3 (108.6)	39 (54)	13.5	17.0
Aju03	Insufficient data	820.2 (54)	1,316.0 (103)	93	92	1,509.0 (284.6)	1,196.3 (108.6)	39 (54)	13.5	24.7
Aju07	n/a	n/a	n/a	n/a	–	n/a	n/a	n/a	n/a	3.9
Aju17	524.1 (89)	7,905.2 (89)	8,373.6 (181)	85	883	3,209.1 (253.9)	931.7 (203.1)	0 (89)	147.1	159.4
Aju18	n/a	n/a	n/a	n/a	–	n/a	n/a	n/a	n/a	38.3
Aju19	n/a	n/a	n/a	n/a	–	n/a	n/a	n/a	n/a	10.9
Aju20	n/a	989.8 (64)	n/a	36	31	49.0 (4.4)	51.8 (6.3)	0 (0)	51.7	81.5
Aju26	590.5 (91)	11,474.1 (91)	n/a	>112	–	n/a	n/a	n/a	n/a	281.0
Aju29	Insufficient data	1,068.5 (46)	1,068.5 (56)	Unknown	–	n/a	n/a	n/a	n/a	9.2
Aju30	577.6 (92)	7,907.3 (92)	9,048.9 (141)	>290	–	n/a	n/a	n/a	n/a	71.0
Aju34	Insufficient data	Insufficient data	Insufficient data	Unknown	–	n/a	n/a	n/a	n/a	24.4
Aju38	265.8 (92)	1,179.9 (92)	3,292.6 (181)	171	418	217.0 (26.9)	121.6 (18.3)	0 (74)	26.1	28.2
Aju40	n/a	n/a	n/a	n/a	–	n/a	n/a	n/a	n/a	4.4
Aju41	147.1 (90)	138.9 (90)	1,099.5 (160)	n/a	–	n/a	n/a	n/a	n/a	78.0
Aju42	n/a	n/a	n/a	n/a	–	n/a	n/a	n/a	n/a	31.9
Aju43	n/a	n/a	n/a	n/a	–	n/a	n/a	n/a	n/a	31.9
Aju44	n/a	n/a	n/a	n/a	–	n/a	n/a	n/a	n/a	31.9
Aju56	250.1 (87)	2,465.0 (87)	13,596.0 (179)	190	497	2,624.8 (94.6)	929.5 (81.5)	0 (0)	118.1	140.6
Aju58	363.0 (91)	2,520.1 (91)	4,694.1 (181)	114	318	176.8 (18.6)	138.7 (13.9)	0 (14)	41.7	59.2
Aju59	n/a	n/a	n/a	n/a	–	n/a	n/a	n/a	n/a	26.0
Aju65	582.6 (91)	1,877 (91)	2,015.0 (165)	13	275	1,614.4 (370.1)	1,182.2 (193.2)	0 (0)	64.7	71.1
Aju66	582.6 (91)	1,877 (91)	2,015.0 (165)	13	275	1,614.4 (370.1)	1,182.2 (193.2)	0 (0)	64.7	65.2

DOI: 10.7717/peerj.1346/table-4

Notes:

a MCP, Minimum Convex Polygon

b M-KDE, Movement-based Kernel Density Estimator

Neither mountain escarpments with 500–600 m elevations (often at near vertical inclines) nor true desert environments limited exploratory movements of cheetahs. Seven adults traversed large areas of the hyper-arid Namib Desert (see Aju30 in Fig. 5). Male Aju17 established a permanent range encompassing sand dunes whilst female Aju29 eventually raised cubs in this environment.

Homing

Despite extensive explorations, only the shortest translocation (Aju65/66 translocated 71 km) resulted in successful homing (Fig. 7). The male coalition returned to the capture property and was assumed to resume its previous home range within two weeks (Fig. 5). All cheetahs translocated >137 km did not home and there was no evidence of oriented, directional movements towards capture sites (Fig. 7). Cheetahs also did not return to the captive facility. The mean distance of last monitoring locations to release sites (Table 4) was 57.1 km ± 63.2 km SD (n = 23 adults) and therefore >14% of the average translocation distance. For non-homing cheetahs (n = 21), the mean distance of last locations to capture sites was 454.1 km ± 197.8 km SD (range: 153.2–826.7 km).

The mean direction of an individual’s last location in relation to release site deviated 173.5° ± 106.2°SD (n = 23) from true home. When analysed in 90°segments around the release centroid, the distribution of directions in relation to true home was random (χ² = 0.13, df = 3, P = 0.988, n = 23), suggesting that cheetahs did not orient toward any particular geographic direction. The distribution of angles located within and outside of the homing sector (i.e., four = in, 19 = out) also significantly differed from the distribution that would be expected if all cheetahs had homed (χ² < 0.01, df = 1, P < 0.0001, n = 23). Neither translocation distance (R² = 0.059, P = 0.269, n = 23) nor time in captivity (R² = 0.012, P = 0.6124, n = 23) were strongly associated with an individual’s post-release orientation.

Range areas

We monitored 14 cheetahs for >90 days and 10 settled into range patterns (Fig. 5), including homing males Aju65/66. Hence, only 35% of released individuals established new ranges. Although Aju29/34 appeared to have settled successfully (Fig. 5), telemetry data were too sporadic to confirm this. The GPS units of Aju26/30 failed before these individuals showed settling behaviour.

Range sizes (excluding Aju65/66) in semi-arid to hyper-arid areas of south-western Namibia varied between 49.0–3,209.1 km² (Table 4). The average distance of geometric range centres (the mean x,y coordinates of range data) to release sites was 66.4 km ± 53.7 km SD (range: 13.5–147.1 km, n = 6). Only the range of Aju01–03 overlapped with the recipient reserve whereas two ranges did not overlap with any PA and four had partial overlap with other conservation areas (Fig. 5). Average percent overlap with all PAs was 38.5% ± 39.0% SD (range: 0–89%) and ranges appeared stable until monitoring ceased.

Site fidelity

Due to extensive explorations, site fidelity was low (Table 5). Mean overlap of daily locations with recipient reserves in year one was 39.8% ± 7.2% SE (n = 23). All cheetahs left recipient reserves and moved into free-hold farmland areas and/or into other private and public conservation areas (Fig. 5). Site fidelity increased significantly when defined to include occupancy of other PAs (Table 5), demonstrating the importance of PA connectivity to improve the safety of far-ranging carnivores released into unfenced environments.

Site fidelity of males and females (Table 5) did not differ significantly when assessed for recipient reserves only (U = 9.5, P = 0.1875) and for all PAs (U = 2.5, P = 0.4102). Likewise, site fidelity did not differ between hard and soft released animals (for recipient reserves and all PAs: U = − 9.5, P = 0.8555). There was also no difference in the mean number of days that hard released (19.0 days ± 15.4 days SD, n = 12) and soft released (19.7 days ± 23.1 days SD, n = 10) cheetahs spent within a 10 km radius around release sites (U = 79.0, P = 0.7069). For soft released individuals, time spent in acclimation pens did not significantly influence site fidelity on recipient reserves (Spearman’s R_s = 0.331, P = 0.3498, n = 10).

Cheetah site fidelity results were also not significantly influenced by recipient area size (R² = 0.0738, P = 0.2099, n = 23). Mean percent overlap of positional data with recipient areas was higher for cheetahs released onto reserves measuring >700 km² (45.7% ± 29.5% SD, n = 13) than for those released onto reserves <350 km² (32.1% ± 40.1% SD, n = 10) but did not differ significantly (U = 117.0, P = 0.0992). This suggests that even the largest private PAs (e.g. NamibRand in Fig. 1) cannot contain extensive post-release movements reliably. Of all cheetahs monitored >90 days (n = 14), eight left recipient reserves permanently whilst six returned sporadically. Neither time in captivity (R² = 0.012, P = 0.6128, n = 23) nor translocation distance (R² = 0.012, P = 0.6129, n = 23) significantly influenced site fidelity results.

Although deployed at near-equal rates (GPS units = 12, VHF/ID units = 11), VHF units provided <27% of positional data in year one. However, the site fidelity estimate for VHF-tagged subjects (recipient reserves only) was more than twice as high as that for GPS-tracked cheetahs (Table 5). We could not detect VHF-tagged cheetahs for prolonged periods (range: 1–253 days), resulting in 124 data gaps in year one (at least one day missing between subsequent locations) with a mean gap interval of 11.0 days ± 34.8 days SD. Conversely, GPS collars produced only 37 data gaps and the mean gap interval (3.4 days ± 4.6 days SD, range: 1–22 days) was significantly shorter (T = 2.36, P = 0.0098). Considering this impact of tracking technology on detection probability, true cheetah site fidelity probably was <25% on recipient reserves, and <47% on all PAs (category GPS in Table 5).

Prey and conflict

Medium to large ungulates comprised the bulk of known wildlife prey (n = 76) (Fig. 8), with springbok (Antidorcas marsupialis) being the preferred prey species (66% of kills). Cheetahs killed approximately similar proportions of sub-adult and adult ungulates (χ² = 1.351, df = 1, P = 0.245, n = 74) and showed no preference for different springbok age classes (χ² = 0.32, df = 1, P = 0.572, n = 50). Female Aju18 hunted as soon as ∼8 h after release, whereas Aju07 did not begin to hunt for 11 days, resulting in her re-capture.

Except for two case studies involved in confirmed livestock predation (the damage was compensated), data sharing with land managers showed that translocated cheetahs caused little conflict (Table 3). During explorations, however, cheetahs frequently roamed across multiple farmland properties within a 24 h period, making effective data sharing impossible. It remains unknown whether they killed livestock during these periods. Farmers from whose properties known livestock raiders had been moved (n = 2) reported new conflict with cheetahs 12–24 months later, showing that translocation was not a permanent solution. Due to continued perceived threats to valuable game (n = 5) and livestock (n = 4), the majority of managers (64%) from whose properties cheetahs were translocated (n = 14) requested repeat removals within two years of first events, giving rise to concerns that sustained translocations could locally induce source sink effects.

Success

Success evaluations were possible for 20 translocated adults (Table 3), giving an overall success rate of 40%. Translocation success was not significantly associated with sex (χ² = 0.84, df = 1, P = 0.3593, n = 20) or release mode (χ² = 1.29, df = 1, P = 0.2526, n = 20) but with degree of habituation to humans (χ² = 6.47, df = 2, P = 0.0394, n = 20). The latter was strongly associated with the amount of time that cheetahs spent in captivity (LogLikelihood = 21.58, χ² = 43.154, df = 2, P < 0.0001, n = 23), suggesting that animals intended for free-range release should not be raised in captivity and ideally be held <250 days (Fig. 9) to increase their chances of survival in anthropogenic landscapes. Spotted hyaena related mortality (n = 4), as another important factor influencing survival, accounted for 33% of confirmed deaths (n = 12) of adults translocated into areas with new competition (n = 18).

All translocations into the far South of Namibia were unsuccessful and were stopped after four early deaths. Conversely, cheetahs in the southwestern pro-Namib ecosystem (n = 12 with known outcomes) and central Namibia (n = 4 with known outcomes) were equally successful (50% respectively). Translocations were carried out at a median cost of $2,760 per cheetah (range: $269–$7,559, n = 23), giving an Individual Conservation Cost of $6,898 per successful adult (see details in Weise, Stratford & van Vuuren, 2014).

Potential cheetah recipient areas in Namibia

Including all parameters to determine recipient area suitability in Namibia (Table 2), our model found no suitable PAs for cheetah releases. Due to extensive exploratory movements (Fig. 5; Table 4), site fidelity had the highest individual parameter impact on area suitability (Table 6), eliminating 95% of available public and private PAs. Currently few unfenced PAs exist that are large enough to prevent cheetahs from re-entering commercial or communal farmlands.

Excluding site fidelity as a prerequisite, CaTSuiT identified 10 isolated patches (Fig. 10) that fulfilled suitability criteria (individual modelling steps in Supplemental Information 4–Supplemental Information 8). Of these, nine patches are <1,800 km² and six patches are <400 km² (Supplemental Information 9), implying that most cheetahs would leave suitable patches within three months of release (see Table 4) and thus be exposed to pre-translocation threats. Even the largest available patch (10,785.3 km² in western Namibia— Fig. 10) was smaller than the observed post-release movements of at least two case studies (Fig. 5; Table 4). In addition, several translocated cheetahs have already moved into the largest predicted patch, and following successful reproduction on nearby reserves (females Aju02/29/58– Table 3), translocations into the pro-Namib ecosystem were stopped in 2012 because an increasing frequency of cheetah observations suggests recovery of the local population (Q Hartung, N Odendaal, S Bachran, pers. comm., 2014).