PROJECT 2
Using camera-trap data to predict abundance and distribution of Cape Leopards
Client: Cape Leopard Trust
Led by: Prof David Borchers (University St Andrews)
Background information
The leopard currently fills the role of apex predator in the Western Cape; however, its conservation status remains uncertain. The species was regularly removed or exterminated from farms with little knowledge of population or genetic status, whether these removals are sustainable or whether the factors giving rise to conflict are established.
Studies show that leopards in the Cape differ morphologically as well as genetically from leopards elsewhere in southern Africa. Leopard home ranges in our study areas may be as much as 10 times larger than those reported in earlier research, suggesting that population numbers may be far smaller than previously estimated.
-
Cape leopards live in the mountains regions of the Cape
-
They like rocky slopes or well vegetated riverine areas rather than open, flat areas
-
Each leopard has unique spot or rosette patterns
-
They move mainly at night but also a little in the day.
Leopards are solitary, moving within a well-defined, sexually exclusive territory or home range.
More background information is at: http://capeleopard.org.za/research/leopard
Project aims and objectives
The Cape Leopard Trust Boland project is a field study of the Cape leopard population in the Boland Mountains. It aims to establish the first rigorous population estimates for leopards in this region, and to identify possible farmer-leopard conflict hotspots. The aims of this project are
-
to investigate the utility of (partly-)automated image recognition software for recapture identification,
- to obtain a baseline estimate of leopard density in the Boland study region, which is necessary for ensuring the survival of leopards in the Cape mountains,
- to estimate leopard distribution in the region, including potential hotspots, and
- to estimate leopard home range size.
About the dataset
The primary data are digital images of leopards taken by the camera traps, together with the location and time that the image was taken. Some example images are shown below.
There are two steps in the analyses of these data:
-
identify recaptures (different photographs of the same animal) and
-
conduct spatial capture-recapture analysis.
Recapture identification: Currently recapture identification is done by eye, but automated image recognition methods are likely to play a central role on many camera trap surveys in the future. In this project students will investigate the utility of the recapture identification software Wild-ID (Bolger et al., 2012: http://software.dartmouth.edu/Macintosh/Academic/Wild-ID_1.0.0.zip) for leopard recapture identification.
Spatial capture-recapture analysis: Given recapture identification, spatial capture-recapture analysis using the maximum likelihood method of Borchers & Efford (2008) of the data will be conducted using the R package secr (Efford, 2015), modelling density as a function of spatial covariates such as those shown in the figures below, which show an elevation map of the study area (right) and land use category (left). Camera trap locations are indicated by black crosses.
Intended outcomes and real-world relevance
Expected outcomes are:
-
an evaluation of the utility of the Wild-ID software for recapture identification, and
-
spatially explicit estimates of leopard density and abundance in the study region, together with estimates of home range size.
Outcome (1) is relevant for the long-term future of the Boland project and the way data are processed in future, while (2) will provide the first spatially explicit estimates for this population, going some way towards meeting the aims of the project.