Inferring patterns of distribution and overlap in carnivores from tourist sightings data

Studying predators is difficult due to low abundance and elusive behaviour, particularly in large savannas like Kruger National Park (KNP). Citizen science presents a promising solution, generating vast datasets across space and time. However, using this data to learn about ecological processes is not without challenges, with key issues being spatial autocorrelation and sampling bias. We used unstructured opportunistic citizen science data from the Latest Sightings© mobile application to obtain ecological insights on predators in KNP. Focusing on four carnivores (lion, leopard, hyena, and cheetah) and one herbivore (impala), we implemented Log Gaussian Cox Process (LGCP) models, incorporating a weighting to account for spatial bias in observations, to examine how species distributions were shaped by vegetation structure, productivity, and proximity to waterways and rest camps. Drivers varying by hunting guild. Lions were strongly attracted to waterways (β=−1.79), while cheetahs exhibited avoidance (β=1.52). Hyenas demonstrated anthropogenic attraction, with relative occurrence increasing near rest camps (β=−0.72). We found significant spatial resource partitioning, particularly between cheetahs and other carnivores. Using predicted distributions, we calculated Bhattacharyya Affinity within a 1km road buffer to quantify spatial overlap among predators. We then examined whether predicted lion-hyena overlap could identify known kill hotspots. A Monte Carlo randomisation test confirmed that observed kill sites were significantly associated with predicted kill hotspots (p<0.05), supporting evidence of kleptoparasitism. Despite inherent biases, our findings are consistent with existing literature on the focal species and highlights the value of unstructured citizen science data, when analysed using appropriate methods, for understanding mammal distributions.