Implications of camera trap survey design and analytical methods for large carnivores estimates

Globally, carnivores face countless threats; and in some cases, these threats are worsened by a lack of population data. In South Africa, conservation has become largely confined to small, fenced, protected areas. It is well established that large carnivores play a vital role in ecosystems, providing valuable ecosystem services, i.e., herbivore regulation and meso-carnivore suppression. If populations are not adequately managed within these small, fenced protected areas, carnivores will place significant pressure on their favoured prey species, and in extreme cases, cause certain species to become locally extinct. For this reason, it is essential for conservation and wildlife managers to understand and monitor large carnivore populations, dynamics and the roles they play within the ecosystems. Estimating population sizes, abundances and densities for species that are difficult to identify (hereafter, unidentifiable), is difficult under conventional capture-recapture methods, leading to a sparse number of estimates on unidentifiable species. The implementation of camera trap methods has been developed and more readily implemented to bridge this knowledge gap, some of which were implemented in this study. Here, I estimated the population sizes of two species, leopard (Panthera pardus) and brown hyaena (Parahyaena brunnea), using camera trap count data from three camera trap projects, analyzed using the package ‘Unmarked’ in R-Studio. Camera trap data was supplied by three previously existing camera trap projects, i.e., Snapshot Safari South Africa, Panthera Organization, and a private baited and non-baited camera trap project. Data supplied were collected across three study sites, namely Madikwe Game Reserve, Pilanesberg National Park, and Atherstone Nature Reserve. The objective of this study was to determine the effect of three camera-trap deployment techniques on space use and density estimates of two large carnivores in Madikwe Game Reserve, Pilanesberg National Park and Atherstone Nature Reserve using unmarked analysis. In this study, I investigated the use of N-mixture models to estimate population sizes of leopard and brown hyaena and how different camera trap deployments influence the N-mixture model population size estimates. I compared N-mixture model population size estimates to pre-existing Bayesian closed-population capture-recapture estimates. Furthermore, this study aimed to provide empirical evidence supporting the use of N-mixture models to estimate the population sizes of both naturally marked and unidentifiable species. This study found that N-mixture models run using data from the sequential baited and non-baited camera trap deployment array and the roadside cluster deployment over-estimated leopard and brown hyaena population sizes across all the study sites. The regular deployment array provided 3 plausible estimates across all three of the fenced protected areas and were closely matched to previous population size estimates. The two targeted approaches, sequential baited and non-baited deployment, and roadside cluster deployment, were more efficient in collecting data. The targeted approaches recorded higher capture numbers and species detection probabilities. The evidence from this research cautions against the use of N-mixture models to conduct population analysis using camera traps due to the model sensitivity, seeing the models are reliant on detection probability and capture numbers.