137 Impacts of heat stress on the accuracy of an ear-tag sensor for detection of eating and rumination behavior in cattle

Abstract Ear-tag accelerometers are used for precision livestock management, but little is known regarding effects of varying environmental conditions on sensor performance. The objective was to evaluate accuracy of a commercially available ear-tag sensor (CowManager SensOor, Agis Automatisering BV, Harmelen, the Netherlands) in quantifying feeding behaviors under varying temperature-humidity indices (THI). RumiWatch System (RW) noseband sensors (Itin+Hoch GmbH, Liestal, Switzerland) were utilized as a benchmark. Nine Holstein-Angus steers (body weight: 488 ± 9 kg) were fitted with CowManager (CM) sensors and RW halters, with continuous monitoring during two 14-d phases: thermoneutral (TN: 21.0 ˚C, 64.0% humidity, THI = 67) and heat stress (HS: cyclical daily temperatures to mimic diurnal patterns; HS: 23.2 ˚C – 33.6 ˚C, 39.5 – 70.0% humidity, THI = 70.3 – 83.5). Steers were individually-housed (3×3 m) and provided ad libitum access to a mixture of corn silage and dried distillers grains with solubles. Hourly summaries of eating time (ET) recorded by CM were compared with EAT1TIME (head down) and EAT2TIME (head up) parameters from RW (v.7.3.36) as well as total ET (EAT1TIME + EAT2TIME); rumination time (RT) was also compared. Agreement was evaluated (R; v.4.3.1) over the full 24-h day (FULL) and a 6-h period from afternoon to early evening (AFT; 1400 – 2000 h) when HS effects would be most pronounced. Across both 14-d phases, agreement between CM and EAT1TIME was moderate to high (concordance correlation coefficient CCC = 0.69 to 0.77), whereas agreement with EAT2TIME and total ET was poor to moderate (CCC = 0.41 – 0.61). Bland-Altman analysis demonstrated significant underestimation of total ET by CM, with biases of -8.98 (-10.88, -7.09) to -6.44 (-7.51, -5.38) min h-1. Agreement for EAT1TIME was high for TN (CCC = 0.74), but only moderate during HS (CCC = 0.69) for the FULL dataset. Similarly, agreement for total ET was moderate for TN but poor for HS over FULL and during AFT, respectively (CCC: TN – 0.53, 0.54; HS – 0.47, 0.41). CowManager significantly underestimated EAT1TIME during HS (bias: FULL = -1.45 -2.60, -0.30; AFT = -2.14 -3.93, -0.35 min h-1), but not TN (bias: FULL = -1.38 -2.82, 0.06; AFT = -1.83 -3.68, 0.02 min h-1). Agreement for RT was high regardless of environmental conditions and time period analyzed (CCC = 0.74 – 0.81). However, during the critical AFT period, CM significantly overestimated RT under HS (bias: 4.42 2.06, 6.79 min h-1) but not TN conditions (bias: -0.47 -2.42, 1.47 min h-1). These results indicate that ET reported by CM is more reflective of time eating with the head down than the total time engaged in eating behaviors. Results of this study also suggest accuracy of ear-tag accelerometers may be impaired when cattle are subjected to heat stress.