Ecologists often study wildlife populations by deploying camera traps. Large datasets are generated using this approach which can be difficult for research teams to manually evaluate. Researchers increasingly enlist volunteers from the general public as citizen scientists to help classify images. The growing number of camera trap studies, however, makes it ever more challenging to find enough volunteers to process all projects in a timely manner. Advances in machine learning, especially deep learning, allow for accurate automatic image classification. By training models using existing datasets of images classified by citizen scientists and subsequent application of such models on new studies, human effort may be reduced substantially. The goals of this study were to (a) assess the accuracy of deep learning in classifying camera trap data, (b) investigate how to process datasets with only a few classified images that are generally difficult to model, and (c) apply a trained model on a live online citizen science project. Convolutional neural networks (CNNs) were used to differentiate among images of different animal species, images of humans or vehicles, and empty images (no animals, vehicles, or humans). We used four different camera trap datasets featuring a wide variety of species, different habitats, and a varying number of images. All datasets were labelled by citizen scientists on Zooniverse. Accuracies for identifying empty images across projects ranged between 91.2% and 98.0%, whereas accuracies for identifying specific species were between 88.7% and 92.7%. Transferring information from CNNs trained on large datasets (“transfer‐learning”) was increasingly beneficial as the size of the training dataset decreased and raised accuracy by up to 10.3%. Removing low‐confidence predictions increased model accuracies to the level of citizen scientists. By combining a trained model with classifications from citizen scientists, human effort was reduced by 43% while maintaining overall accuracy for a live experiment running on Zooniverse. Ecology researchers can significantly reduce image classification time and manual effort by combining citizen scientists and CNNs, enabling faster processing of data from large camera trap studies.

No abstract

Food caching is a common strategy used by a diversity of animals, including carnivores, to store and/or secure food. Despite its prevalence, the drivers of caching behaviour, and its impacts on individuals, remain poorly understood, particularly for short-term food cachers. Leopards Panthera pardus exhibit a unique form of short-term food caching, regularly hoisting, storing and consuming prey in trees. We explored the factors motivating such behaviour among leopards in the Sabi Sand Game Reserve, South Africa, associated with four not mutually exclusive hypotheses: food-perishability, consumption-time, resource-pulse and kleptoparasitism-avoidance. Using data from 2032 prey items killed by 104 leopards from 2013 to 2015, we built generalized linear mixed models to examine how hoisting behaviour, feeding time and the likelihood of a kill being kleptoparasitized varied with leopard sex and age, prey size and vulnerability, vegetation, elevation, climate, and the immediate and long-term risk posed by dominant competitors. Leopards hoisted 51% of kills. They were more likely to hoist kills of an intermediate size, outside of a resource pulse and in response to the presence of some competitors. Hoisted kills were also fed on for longer than non-hoisted kills. At least 21% of kills were kleptoparasitized, mainly by spotted hyaenas Crocuta crocuta. Kills were more likely to be kleptoparasitized at lower temperatures and if prey were larger, not hoisted, and in areas where the risk of encountering hyaenas was greatest. Female leopards that suffered higher rates of kleptoparasitism exhibited lower annual reproductive success than females that lost fewer kills. Our results strongly support the kleptoparasitism-avoidance hypothesis and suggest hoisting is a key adaptation that enables leopards to coexist sympatrically with high densities of competitors. We further argue that leopards may select smaller-sized prey than predicted by optimal foraging theory, to balance trade-offs between kleptoparasitic losses and the energetic gains derived from killing larger prey. Although caching may provide the added benefits of delaying food perishability and enabling consumption over an extended period, the behaviour primarily appears to be a strategy for leopards, and possibly other short-term cachers, to reduce the risks of kleptoparasitism.

Human impact is near pervasive across the planet and studies of wildlife populations free of anthropogenic mortality are increasingly scarce. This is particularly true for large carnivores that often compete with and, in turn, are killed by humans. Accordingly, the densities at which carnivore populations occur naturally, and their role in shaping and/or being shaped by natural processes, are frequently unknown. We undertook a camera-trap survey in the Sabi Sand Game Reserve (SSGR), South Africa, to examine the density, structure and spatio-temporal patterns of a leopard Panthera pardus population largely unaffected by anthropogenic mortality. Estimated population density based on spatial capture-recapture models was 11.8 ± 2.6 leopards/100 km 2 . This is likely close to the upper density limit attainable by leopards, and can be attributed to high levels of protection (particularly, an absence of detrimental edge effects) and optimal habitat (in terms of prey availability and cover for hunting) within the SSGR. Although our spatio-temporal analyses indicated that leopard space use was modulated primarily by "bottom-up" forces, the population appeared to be self-regulating and at a threshold that is unlikely to change, irrespective of increases in prey abundance. Our study provides unique insight into a naturally-functioning carnivore population at its ecological carrying capacity. Such insight can potentially be used to assess the health of other leopard populations, inform conservation targets, and anticipate the outcomes of population recovery attempts. K E Y W O R D Scarnivore ecology, carrying capacity, Panthera pardus, population regulation, spatial capture-recapture