Predicting cattle grazing behavior on rangeland using accelerometers

Abstract The objective was to determine if 3-axis accelerometers could be used to predict daily activity for cattle grazing rangeland. There were 48 Hereford × Angus 2-yr-old low- or high-residual feed intake (LRFI or HRFI) cows used in this 2-yr trial. Cattle grazed in 4 pasture treatments consisting of continuously grazed, control (CCON); continuously grazed, supplemented (CTRT); rotationally grazed, control (RCON); and rotationally grazed, supplemented pastures (RTRT). Three LRFI- and 3 HRFI-collared cows in each treatment had accelerometers mounted for 29 d in 2016 and 45 d in 2017, beginning mid-October. Grazing time (GT), resting time (RT), and walking time (WLK) were obtained for each cow by direct observation over 3 d each year and compared with accelerometer predicted behavior. In 2016, 1.6% of the days were rejected for halter-mounted accelerometers and 3.6% were rejected in 2017 for collar-mounted accelerometers. The GT and RT were more accurately predicted than was WLK with the percentage error of predicted against observed data being 11.94% for RT, 13.51% for GT, and 30.13% for WLK in 2017. Less observation data were available in 2016, but when considering other sampling periods for the same cows and halters, the error rate was 15.1% for RT, 19.3% for GT, and 52.6% for WLK. The accelerometers successfully identified patterns of grazing behavior and differentiated among climatic, grazing system, supplementation status, and residual feed intake classification influences on GT, RT, and WLK. In a more moderate climate year, HRFI cattle appeared to rest less (P