Development of a predictive model to identify the day of lambing in extensive sheep systems using autonomous Global Navigation Satellite System (GNSS)

The aim of this project was to develop a model capable of predicting day of lambing from GNSS data. A field trial was conducted in New Zealand over a two-week period from September to October 2017. Forty ewes were fitted with GNSS tracking collars, recording at three-minute intervals. Animals were visually observed to record birth events. Of the 40 ewes, 25 lambed during the experimental period. A heuristic model was developed to predict the day of lambing by classifying ewes as either 'non-lambing' or 'lambing' on each day of the study based on her speed of movement, mean distance to peers and extent of spatial landscape. Predictions were based on the series moving over or under a given threshold, determined through estimated least-square means from the linear mixed-effects statistical analysis. The overall accuracy of the model was 83.0%, with a sensitivity of 63.6% and specificity of 84.1%. The results of this project highlight the potential for GNSS tracking for post-hoc detection of lambing events. This could be used to inform graziers on the status of individual animals and their flock as a whole, particularly in extensive farming systems where close observation may be limited. Further work should be conducted on larger datasets to improve reliability of results.