Potential of Accelerometers and GPS Tracking to Remotely Detect Perennial Ryegrass Staggers in Sheep

• Published in: Smart agricultural technology 2022-12, Vol.2, p.100040, Article 100040
• Main Authors: Trieu, Ly Ly, Bailey, Derek W., Cao, Huiping, Son, Tran Cao, Scobie, David R., Trotter, Mark G., Hume, David E., Sutherland, B. Lee, Tobin, Colin T.
• Format: Article
• Language: English
• Subjects: Accelerometer; Machine learning; On-animal sensor; Perennial ryegrass staggers
• ISSN: 2772-3755; 2772-3755
• DOI: 10.1016/j.atech.2022.100040

•Ryegrass staggers is a neurotoxic condition that adversely affects livestock•Visual detection of ryegrass staggers is labor intensive•Accelerometers can detect behavior changes associated with ryegrass staggers•Machine learning and accelerometers have potential to detect ryegrass staggers•Random Forests was superior to Signal Vector Machines for prediction of staggers Perennial ryegrass staggers (staggers) is a neurotoxic condition in livestock that is caused by consumption of ryegrass (Lolium perenne) infected with specific strains of Epichloë fungal endophytes. These grass-endophyte associations produce toxins that can adversely affect animals and can in some cases lead to death. In sheep, symptoms typically include head shaking, changes in gait, stiffness and falling. Affected sheep can recover after removing them from pastures containing toxic strains of endophyte. A pilot case study was conducted in Lincoln, New Zealand to determine if ryegrass staggers could be identified with data collected through GPS tracking and accelerometers. Fourteen sheep per treatment grazed in either a toxic endophyte-infected ryegrass paddock or an endophyte-free control paddock for 17 days in late March and early April 2017. Randomly selected sheep were fitted with collars containing a 3-axis accelerometer recording movements at 12 Hz (10 collars in endophyte infected paddock and 6 in the control paddock). Three sheep per treatment were also tracked at 3-minute intervals with GPS receivers. Sheep were scored by an experienced observer for symptoms of staggers weekly and at the end of the study using a 0 to 5 scale. Control sheep did not display any symptoms of staggers and 10 sheep in the infected pasture displayed little or no symptoms (0 or 1 score). The other 4 sheep in infected pasture had scores from 2 to 4 at the end of the study. Sheep grazing in the infected pasture (2.91 m/min ± 0.04 SE) moved slower (P=0.04) than sheep in the control pasture (3.12 m/min ± 0.05 SE). Distance travelled varied among days, but there did not appear to be any temporal trends. Machine learning analyses of accelerometer data showed that the behavior of affected sheep changed during the study. Activity of sheep displaying symptoms (scores ≥ 2) increased more in the morning and midday during the latter part of the study than control sheep and sheep with few or no symptoms (score < 2). However, behavior of individual sheep at night remained relatively consistent during the study. Accelero