Reassessment of an introduced cheetah Acinonyx jubatus population in Matusadona National Park, Zimbabwe

Abstract Translocations are used to mitigate human–wildlife conflict, secure population viability of isolated populations and introduce or reintroduce populations in former or new range. With wild species increasingly confined to small patches of habitat embedded in human-dominated landscapes, the use of translocations is likely to increase. The cheetah Acinonyx jubatus is a large carnivore species with a long history of translocations. As for other species, evaluation of the success of cheetah translocations is complicated by a scarcity of published results, especially of failed attempts. Yet, such information is crucial to improve future translocations. A relatively well documented case is the translocation of alleged problem cheetahs into Matusadona National Park, Zimbabwe, in the early 1990s. In this study we used a combination of survey methods to reassess the status of Matusadona's cheetah population and model current occupancy in relation to densities of competing carnivores and altitude. Our findings indicate this cheetah population has effectively been extirpated, highlighting the importance of thorough planning and standardized long-term monitoring of translocated populations for the understanding of the factors that affect translocation success.


Introduction
A s a consequence of increasing human populations and the resulting demand for agricultural land and pressure on natural resources, the ranges of most large carnivores have contracted substantially (Di Minin et al., ).Maintaining viable populations of large carnivores requires large connected habitat networks (Di Minin et al., ).
However, such networks are often inhabited by people, who degrade carnivore habitat and persecute large carnivores because of actual or perceived livestock losses, confining them to isolated patches of land (Di Minin et al., ).
Translocation is regularly used to mitigate conflict, secure viability of isolated populations and reintroduce populations in former or new range (Linnell et al., ; Fischer & Lindenmayer, ).Success depends on the aim of the translocation, the source of the translocated individuals, the number of individuals released and the cause of the original decline (Fischer & Lindenmayer, ).Translocations generally fail to reduce human-carnivore conflict and it is now widely accepted that such negative interactions are best mitigated by promoting coexistence (Linnell et al., ; Fischer & Lindenmayer, ; Boast et al., a).However, translocations to establish or re-establish and augment populations of native, non-vulnerable, wildlife species seem more successful (Griffith et al., ).Translocation success increases when the source population is wild, the founder population is large (Griffith et al., ; Fischer & Lindenmayer, ) and the original cause of the population decline has been eliminated (Fischer & Lindenmayer, ).Although it is recognized that monitoring is crucial to improve the success of translocations (IUCN/SSC, ), there has been little follow-up of translocated animals to determine success (Linnell et al., ) and long-term evaluation of introduced or reintroduced populations is often lacking (Fischer & Lindenmayer, ).
The cheetah is a large carnivore species regularly translocated for conservation purposes (Boast et al., a).A relatively well-documented translocation is the - translocation of alleged problem cheetahs from commercial farmland into Matusadona National Park, Zimbabwe, an unfenced protected area without resident cheetahs (Zank, ; Pitman, ).Although it was initially thought the introduced cheetah population would not be viable without intensive management (Zank, ), it persisted without intervention and, with the establishment of a breeding population, the introduction was cautiously deemed a success (Purchase & Vhurumuku, ).Whether the translocation reduced human-cheetah conflict on commercial farmland is unknown (Purchase, ).In this study, we used citizen science, questionnaire-based information, camera-trap and spoor surveys to reassess the status of Matusadona's cheetah population and model current cheetah occupancy in relation to densities of competing carnivores and terrain elevation.Our follow-up highlights the importance of thorough planning and standardized long-term monitoring of translocated populations to understand the factors affecting translocation success and to improve future translocation programmes.

Study area and species
The , km  unfenced Matusadona National Park, utilized for photographic safaris, lies in northern Zimbabwe on the shore of Lake Kariba.Depending on the lake level, the Park comprises a valley of c.  km  and an escarpment of c. , km  (Fig. ).During -,  alleged problem cheetahs were opportunistically removed from commercial farmland in southern Zimbabwe; six died shortly after capture, and  were kept in a captive enclosure (boma) for  weeks and then soft released into the Park, which held no resident cheetahs at the time (Zank, ; Pitman, ).Although it was assumed cheetahs were once resident in Matusadona, there are no records to support this (Zank, ) and it remains unknown whether cheetahs had become extinct or had never been present at all (Purchase, ).The introduced cheetahs utilized only the valley floor (Purchase, ), which has a grassland foreshore but largely consists of woodland with Colophospermum mopane, Combretum spp.and Terminalia spp.

Data collection
As part of a nationwide survey (van der Meer, ), we launched a citizen science programme (-present) to collect cheetah sightings and photographs from tourists, safari operators and others.In addition, we interviewed  respondents (August-September ) about cheetah presence in Matusadona and surrounding areas:  Zimbabwe Parks and Wildlife Management Authority rangers, seven CAMPFIRE rangers, six safari guides, five hunters and eight hunting scouts (van der Meer, ).If respondents had seen cheetahs, we recorded sighting date and location, number of individuals, age, sex and size.We only included sightings from respondents who could identify the cheetah from carnivore pictures and clearly recalled sighting details.Some respondents provided photographs with their sightings.We also asked respondents about human-cheetah conflict and anthropogenic cheetah mortality, and checked reports at the Zimbabwe Parks and Wildlife Management Authority and CAMPFIRE offices for records of conflict and mortality.During June-August  we carried out a camera-trap survey using a pre-planned grid of  traps set up - km apart on the valley floor and part of the escarpment (Fig. ).Each trap station consisted of two Cuddeback (Cuddeback, Wisconsin, USA) or Panthera V (Panthera, New York, USA) cameras, c.  m apart, facing each other but slightly offset, to limit interference from the flash.The cameras were attached to trees or metal standards c.  cm above ground level.Following establishment, the survey was run for  days, after which traps were removed within the same time frame and order in which they were deployed.
In June  we drove three road transects on the valley floor (total length . km), and in July-August  we drove four road transects on the valley floor (total length . km), all of which were driven three times in the early morning, on different days, and three of which repeated routes surveyed in .Transects were driven at - km/h, with an experienced tracker on a seat mounted at the front of the vehicle.When large mammal spoor was located the vehicle stopped and spoor was identified and date, location (with a GPS) and the number of animals recorded.To avoid double counting, transects were never surveyed on consecutive days and we only recorded tracks ,  hours old.
We consulted expert authorities and literature to gain insight into historical cheetah population trends in Matusadona and possible underlying causes of changes in cheetah numbers; e.g.fluctuations in lake levels, which affect the area of valley floor foreshore exposed and therefore prey and predator densities.After  there have been no prey surveys other than aerial counts.Therefore to facilitate a comparison between studies we used dry season aerial count data with comparable transects as an indication of impala and buffalo densities.Although aerial counts have limited detection ability, they provide an indication of population trends (Dunham et al., ).From  lion and spotted hyaena densities were based on spoor surveys rather than on direct sightings or call-ups (playback recordings of vocalizations).

Analyses
From the photographs collected via sightings (n = , % with photographs) and the camera-trap survey (n = ), we identified individual cheetahs based on coat markings (Zank, ).Sightings of known individuals (n = ) were used in a presence-absence model and to determine space use based on minimum convex polygons (MCPs).Although MCPs create unpredictable bias when sampling is not systematic, as with opportunistic sightings, (Börger et al., ), we nevertheless decided to use this method, to facilitate comparison between studies.
Substrate and cheetah density in Matusadona were sufficient to use the calibration model of Winterbach et al. (), in which observed spoor density = .× carnivore density.In the absence of a species-specific equation based on a number of populations, this is the recommended calibration equation for the cheetah (Boast et al., b).For each transect we calculated spoor density (individuals per  km).Mean spoor density, standard deviation and % confidence intervals were calculated across all transects driven.Based on these calculations we determined mean, minimum and maximum cheetah numbers for the valley floor.
For the camera-trap survey, the MCP encompassing all traps buffered by  km was gridded at  ×  m resolution.This gridded habitat mask was used to extract the spatially explicit recapture modelled densities for sympatric carnivores and altitude (GLOBE Task Team, ).In addition, total numbers of cheetah sightings were calculated per grid cell by counting the presences () and absences () of identified individuals.Given the number of zeros in the data, a zero-inflated Poisson generalised additive model was applied to model the probability of occurrence of cheetahs in relation to altitude and densities of competing carnivore species using zipGAM in package mgcv in R .. (Wood, ; R Core Team, ).

Population estimate, movement and range
Based on photographic identification, the cheetah population of Matusadona comprised three adult cheetahs: a female (first sighting September , last sighting May ) and a coalition of two males (first sighting December , last sighting May ).We did not receive photographs of other individuals, nor did sightings without photographs indicate the existence of additional cheetahs or breeding.The analyses of spoor survey data indicated a mean population size of - cheetahs (Supplementary Table ).We only encountered spoor of adult cheetahs, on their own or in a group of two individuals.Cheetahs predominantly utilized the north-eastern part of the valley floor; they were not sighted on the escarpment or in the neighbouring communal lands.Home ranges overlapped by -% (c.. km  ) and were . km  for the cheetah female and . km  for the male coalition (Fig. ), which is larger than the .-.km  home ranges in  (Purchase, ).The cheetah female was last seen in May  and is presumed dead.

Factors affecting current occurrence
The presence-absence model indicated that terrain is the strongest explanatory variable for cheetah occurrence.Cheetah presence was restricted to below  m altitude (Fig. ).Cheetah presence was also related to the occurrence of competitors: although spotted hyaena densities did not significantly influence cheetah presence, both lion and leopard densities did.As lion and leopard (. . individuals/  km  ) densities increased, the likelihood of cheetah occurrence decreased (Fig. ; Supplementary Table , Supplementary Fig. ).The model explained % of the deviance in cheetah presence/absence and, based on altitude and densities of competing predators, predicted cheetah occurrence to be higher along the foreshore of the valley floor (Fig. ).

Population trends and mortality
Our estimate of the cheetah population from spoor surveys corresponded well with the estimate from sightings, but in the study of Purchase & Vhurumuku () the estimate based on spoor density was substantially higher than the estimate based on sightings.To facilitate comparison we therefore only present estimates based on sightings.During - Matusadona's cheetah population reproduced and remained relatively stable at a mean of . ± SE . individuals (Table ).However, the proportion of subadult cheetahs increased, indicating lower juvenile mortality but possible higher adult mortality (Table ).Data on survival rates of cubs and adults are scarce: in  mean litter size was . cubs (n = ) and cub survival at - months was % (Purchase, ).In this same study, it was confirmed that two cheetahs of the founder population were still alive.Matusadona is surrounded by communal land and opportunities for cheetahs to disperse are limited (Zank, ).Although human-cheetah conflict in the communal land surrounding Matusadona has historically been minimal (one report in ; Purchase & Vhurumuku, ), cheetahs ranging into this communal land have suffered from anthropogenic mortality (Zank, ).We did not record any cases of human-cheetah conflict or reports of anthropogenic mortality.

Ecological factors affecting population trends
Shortly after the cheetah introduction, the lake level increased, with a peak in  and high levels until  (Fig. ), decreasing the area of foreshore exposed.Aerial surveys suggest this negatively affected the impala population, but this did not seem to impact the cheetah population.Within this period the cheetah population fared well, which was attributed to a decline in lion numbers (Purchase & Vhurumuku, ).However, despite lion numbers remaining low (Fig. ), during - the cheetah population decreased dramatically (Table ).Within this period the lake level decreased, reaching its lowest point in  (Fig. ), exposing a larger area of foreshore, which coincided with high impala numbers (Fig. ).Disease-related mortality of wild cheetahs is minimal, but cheetahs are susceptible to anthrax (Terio et al., ).Since  there have been no records of anthrax outbreaks (Department of Livestock and Veterinary Services, pers.comm.), nor did we come across records of sudden simultaneous die-offs of cheetahs or other carnivores in Matusadona.

Discussion
Based on our survey and recent citizen science sightings, the Matusadona cheetah population currently comprises a coalition of two male cheetahs.This population is considered isolated (see also Zank, ; Purchase, ) and, with only two males . years old remaining, is functionally extirpated.The reasons for the drastic decline of this cheetah population remain unclear.Aerial surveys suggest that a reduction in productive habitat as a result of high lake levels negatively affected prey availability, which may have resulted in a decrease in cheetah numbers.However, prey estimates based on road transects and block counts do not support this hypothesis (Purchase & Vhurumuku, ).Post-release population viability analyses showed a viable cheetah population could be established if cub mortality was # % (Zank, ).It was initially expected that cub mortality would be higher (%) and that intensive metapopulation management would be necessary to maintain the population, but later studies showed cub mortality was % and the population would be viable at a carrying capacity of $  cheetahs (Purchase, ).
Whether Matusadona could ever sustain this number of cheetahs is questionable.Even though the territories of Matusadona's male cheetahs overlapped considerably (Zank, ; Purchase, ), territorial behaviour in combination with limited availability of preferred habitat is likely to restrict the number of cheetahs that can successfully utilize the  km  valley floor (see also Purchase, ).This limits the carrying capacity of the Park which, based on historical sighting-based population sizes, is likely to be ,  cheetahs.
Intraspecific competition with larger carnivores affects cheetah densities (Lindsey et al., ) and, apart from the availability of optimal habitat and prey, impacts cheetah occurrence on the valley floor.Our presence-absence model showed no significant effect of spotted hyaena densities but a negative relationship with lion and leopard densities.Despite signs of leopard predation of cheetahs (Purchase, ), previous studies in Matusadona focused on competition with lions and spotted hyaenas; the potential impact of leopards was not taken into account.Although lions and spotted hyaenas are considered the main competitors of cheetahs (Caro, ), leopards have considerable dietary overlap with cheetahs (Lindsey et al., ) and cause cheetah mortality (Caro, ; Broekhuis, ).Consequently, cheetahs actively avoid interactions with leopards (Vanak et al., ) and the presence of leopards increases the cheetah's space requirements (Lindsey et al., ).
For free-ranging mammal populations in which individuals mature late and have few offspring per year, adult and juvenile survival rates contribute more to population growth than fecundity (Heppell et al., ).Although Matusadona's .% adult and subadult mortality rate was considered low (Purchase, ), the mortality rate of dispersers is likely to be considerably higher.Cheetahs that moved into neighbouring communal land were not resighted (Zank, ) and there are no indications that Matusadona's cheetahs colonized nearby wildlife areas or connected to other resident cheetah populations (van der Meer, ).Since the release, human population density (ZPWMA, ) and numbers of livestock in the region have increased, and numbers of wild prey decreased (Dunham et al., ), thereby further reducing dispersal potential for cheetahs (Winterbach et al., ).As suggested by Purchase (), the viability of Matusadona's cheetah population is likely to have been impaired by high mortality rates of subadults, which were forced to disperse into suboptimal habitat by already established cheetahs.
The Matusadona cheetah introduction was based on assumptions rather than a feasibility study to assess whether the cheetah's ecological needs could be fulfilled (Zank, ; Pitman, ).Twenty-one cheetahs were captured opportunistically without considering factors such as relatedness and ideal age and sex-ratio for introduction (IUCN/SSC, ).In addition, this number of removals and translocations was unlikely to significantly reduce human-cheetah conflict at the capture site (Purchase, ) nor was it sufficient to establish a viable population at the release site (Zank, ).If international guidelines had been followed (IUCN/SSC, ), the translocation should not have proceeded, especially as without any attempt to identify and mitigate factors causing low cheetah densities inside protected areas and declining cheetah numbers outside protected areas, the translocation was unlikely to significantly contribute to cheetah conservation in Zimbabwe (Zank, ).
Long-term protocols in which key parameters are monitored at previously specified time intervals are crucial to understand the factors affecting translocation success and improve future translocations (Fischer & Lindenmayer, ; IUCN/SSC, ).Compared to other translocated populations (Boast et al., a), the Matusadona cheetah population is relatively well studied but, nevertheless, a  Purchase ();  citizen science sightings ( identified cheetahs,  non-identified cheetahs). Purchase & Vhurumuku ();  citizen science sightings in  and  citizen science sightings in  (some identified cheetahs, some non-identified cheetahs);  citizen science sightings during April -April . Present study:  camera-trap and citizen science sightings accompanied by photographs (identified cheetahs).lack of standardized long-term monitoring prevents us from determining the exact causes of the population's extirpation.Translocation success largely depends on habitat quality and quantity at the release site and, in the case of translocations in unfenced environments, the surrounding areas (Griffith et al., ; Fischer & Lindenmayer, ; Boast et al., a).Although the Matusadona cheetah population persisted for over a decade, the limited availability of optimal cheetah habitat in combination with interspecific competition and minimal dispersal abilities appear to make the area unsuitable to harbour a viable cheetah population.Unless the exact causes for the extirpation of Matusadona's cheetah population can be clearly understood and properly addressed, further reintroductions (IUCN/SSC, ) of cheetahs into the area are not advisable.
The escarpment is dominated by miombo woodland and is characterized by steep valleys with limited water availability in the dry season (Zank, ; Purchase, ).The cheetah's competitors, lion Panthera leo, leopard Panthera pardus and spotted hyaena Crocuta crocuta, occur in the Park.For Matusadona's large carnivores, buffalo Syncerus caffer and impala Aepyceros melampus have historically been the most abundant prey species (Zank, ; Purchase, ).The Park is bordered by the Sanyati and Ume rivers and surrounded by communal land used for subsistence farming and trophy hunting under the Communal Areas Management Programme for Indigenous Resources (CAMPFIRE; Fig. ).

FIG. 1
FIG. 1 Matusadona National Park with positions of camera-trap stations and spoor survey transects for the cheetah Acinonyx jubatus, and the contours of the shoreline at high ( m) and low ( m) lake levels (shoreline contours from Purchase, ).

FIG. 2
FIG. 2 Home ranges of the three identified cheetahs in Matusadona National Park.
Cheetahs select home ranges based on prey availability and vegetation characteristics, and within these home ranges they preferably utilize open savannah habitat (Caro, ; Broomhall et al., ).Matusadona's cheetahs only utilized the flat valley floor, within which they made extensive use of a small core area of open habitat along the north-eastern lake shore.Regardless of lake level, this area has the largest foreshore (Fig. ; Purchase & Vhurumuku, ), highest impala density (Zank, ; Dunham et al., , ) and optimal habitat to hunt and rest (Purchase, ).Although cheetah females tolerate each other, males actively defend territories (Caro, ).

FIG. 3
FIG. 3 Estimated smooth functions (s) for the selected covariates (a) lion density; (b) leopard density; (c) altitude and (d) hyaena density, in the occurrence model for cheetahs in Matusadona National Park.The solid line is the estimated smooth function, shaded regions are the approximate % confidence intervals and on the y-axis is the smoothed function of the number of cheetah individuals and respective estimated degrees of freedom for the model terms.Cheetah numbers' distribution is displayed as the rug plot (small lines) along the x-axis.

TABLE 1
Overview of the historical and current size and demography of Matusadona's introduced cheetah Acinonyx jubatus population.