On-animal sensors may predict paddock level pasture mass in rotationally grazed dairy systems

Abstract

Precision livestock farming aims to improve animal welfare and farm management using digital technology. We investigated the potential of individual on-animal sensors to predict paddock-level pasture mass, an important metric for grazing management in pasture-based dairy systems. The study consisted of four groups of 25 cows assigned to different pasture allocations (ranging from an estimated 80% to 120% of their energy requirements) over two 20-day experimental periods (late spring and late summer). Each cow was fitted with five sensors that measured a range of behaviours, including rumination time, eating/grazing time, activity and lying time. These data were used to build predictive models of pasture mass, which was estimated by calibrated rising plate meter. Our results show that rumination time was the most critical behaviour for predicting paddock-level pasture mass; the best predicted was post-grazing pasture mass (kg DM/ha) with a maximum Adjusted-R² value of 0.58 using a linear model. Including pasture and behaviour data at less than 24-hour resolution did not improve model performance, likely due to the importance of rumination, which is a diurnal behaviour. It is unclear whether this level of predictive ability is practically useful for making grazing management decisions; however, given its near real-time nature, low effort, and objectivity, the approach may provide value to farmers. Further evaluation is needed to determine how providing these data affects farmers' decision-making processes and therefore its value. In conclusion, our proof-of-concept experiment demonstrates the potential of individual on-animal sensors to predict post-grazing pasture mass, and this could help farmers make informed decisions for grazing management.