Quantifying Sheep Behaviour Using a 3D Accelerometer: A Proof-of-Concept for Objective Stress Assessment

Continuous digital monitoring of sheep behaviour shows potential for early stress detection. In Part 1 of this study, a novel accelerometer-based behaviour-recognition system using a nRF52832 microcontroller with Bluetooth wireless data transfer was developed and validated. A dedicated algorithm was developed to focus on the automatic detection of rumination, which also enables the classification of resting/idling and eating. The system achieved accuracies of 0.87 (rumination), 0.90 (resting/idling), and 0.86 (eating). Specificities were 0.87, 0.95, and 0.94; sensitivities 0.89, 0.80, and 0.60; and precisions 0.79, 0.88, and 0.73, respectively. In Part 2, four sheep were continuously monitored for 24 h to establish baseline behavioural durations. Animals were then relocated in pairs to an unfamiliar enclosure for a further 24 h observation period. Relocation resulted in a significant reduction in rumination time (-45.6%, <i>p</i> < 0.05) and a significant increase in resting/idling (+47.9%, <i>p</i> < 0.05), while time spent eating decreased but did not reach statistical significance (-36.2%). These findings indicate that detecting deviations from baseline rumination and resting/idling durations may serve as suitable ethological parameters for automated, sensor-based stress alerts. With further technical refinement and validation, the developed system shows strong potential as a reliable, non-invasive tool for monitoring key sheep stress indicators.