Lambing detection in East Friesian dairy sheep based on area-specific acoustic and movement activity monitoring

Lambing is a critical period in sheep husbandry, and dystocia is a major contributor to ewe and lamb mortality. Current monitoring practices rely largely on manual observation, which is labor-intensive and may not allow timely intervention. In this proof-of-concept study, we evaluated whether the onset of lambing can be inferred from pen-specific acoustic and movement activity. Behavior monitoring devices equipped with beamforming microphone arrays and radar modules were installed above individual lambing pens and used to monitor 12 East Friesian ewes housed at a research facility. Data were evaluated as daily differences in hourly aggregated raw data using classical change-detection methods for time-series data that can be applied in real time with minimal computational power. Acoustic activity, which is the number of acoustic events per pen and time interval, increased significantly toward the hour of lambing and enabled prediction of lambing 1 h in advance (AUC = 0.87). Movement activity, defined as the number of movements per pen and time interval, also increased before and during the hour of lambing but showed lower discriminative performance (AUC = 0.77) and a higher false-positive rate. The combination of both parameters resulted in a detector with an AUC of 0.91, corresponding to 89% sensitivity at 88% specificity. These findings indicate that barn-installed, pen-level monitoring of acoustic and movement activity is a feasible, low-effort alternative to animal-attached sensors for lambing detection. Such a system could alert farmers to impending parturition, enabling timely assistance and potentially reducing perinatal losses. Future work should validate its performance under commercial conditions across farms with larger flock sizes and diverse breeds and include a comprehensive assessment of economic viability.