Modelling the predation sequence of wolves hunting sheep: Implications for evaluating the effectiveness of preventive methods

Abstract

Although accounting for the sequential nature of the predation is useful to better understand predator-prey interactions, this approach has rarely been adopted in the context of large carnivore attacks on livestock. Here we developed a model of wolf (Canis lupus) predation that integrated its sequential nature and the effects of covariates. We used over 300 wolf observations obtained by thermal infrared cameras in hotspots of wolf predation on sheep in the French Alps to quantify the impact of the presence of livestock guarding dogs (LGDs), fencing and other environmental factors on the conditional probabilities of wolf approach and attack. Our data show that most often, only one or two wolves appear to be involved in predation on sheep, suggesting that the predator is unlikely to overwhelm the LGDs by sheer numbers. However, we detected no effect of LGDs or fencing on the predation sequence, possibly because observations of wolf behaviour were brief and subject to high variability among observers and years. Observers did not systematically record data and sometimes scored contradictory information. A posteriori power analysis revealed that the effects of LGDs or fencing would have been detected in our data if they halved and one-third the probability of approach and attack, respectively. We cannot draw general conclusions about the effectiveness of LGDs or fencing because our data were not random but collected in hotspots only. Nevertheless, our approach can be used to improve assessment of the effectiveness of preventive methods, a vital task for the conservation of large carnivores.