Abstract. GPS technologies represent a valuable tool to study animal behaviour remotely. The research is aimed at evaluating the use of GPS collars equipped with activity sensors to infer behaviour in grazing cattle. Six cows from two breeds (Limousin, Chianina) were equipped for 8 months with GPS collars with built-in tri-axial accelerometers providing activity measurements every 152 seconds. Four behaviours (resting, rumination, feeding, walking) were directly observed in synchrony with collar measurements. Behaviours were classified with discriminant analysis (DA), considering a set of six predictors and their logarithm: movement rate (metres/hour), activity measurements on the X (forward/backward) and Y (sideways) axes and their difference, mean and variance. We tested several models and reported the ones with the highest rate of correct classifications. These were achieved by adopting separate models for the two breeds and, within breed, for two season-related periods (spring/summer and autumn). This suggests that the environmental effects (e.g., weather), as well as the breed-specific habits, have to be taken into account when inferring behaviour in grazing animals, since they produce significant alterations in intensity of activity. Walking activity was misclassified in more than 80 % of cases, while rumination and resting behaviour were mutually mistaken. Feeding and walking were thus merged to obtain active behaviours, and rumination and resting were classified as inactive behaviours. On average, DA correctly classified over 90 % of active intervals and 85 % of inactive behaviours. In conclusion, the simultaneous use of GPS and activity sensors represents a useful technique to track movements of grazing livestock and, at the same time, to discriminate between active and inactive behaviours. This information could provide benefits for rangeland management in terms of improving their efficiency of utilisation and enhancing the productive performances of animals.Keywords: accelerometer, GPS/GSM collars, activity budget, classification, Gaussian Finite Mixture Models.
IntroductionThe study of behaviour in grazing livestock assumes an essential role in research topics related to rangeland ecology and management, as well as in animal husbandry practices. The assessment of sustainable grazing systems, aiming at the mitigation of negative impacts deriving from excessive or reduced grazing, such as biodiversity loss [1], requires specific knowledge of plant ecophysiology associated with herbivore grazing behaviour and dynamics [2]. Thus, understanding the drivers of resource selection and mechanisms adopted by animals in order to cope with environmental conditions is crucial in free-ranging systems [3]. In livestock intensive farming, the continuous monitoring of behaviour provides farmers the opportunity to infer health conditions and well-being of animals, as well as to improve their management and enhance animal productive performances [4].The possibility to implement an efficient system for precision monitor...