85. Evaluation of Automated Feed Behavior Sensors for Intake Prediction in Growth-phenotype Divergent Crossbred Angus Beef Steers

Abstract In an experiment evaluating the effect of phenotypic selection based on sire yearling weight expected progeny difference, CowManager ear sensors (Agis Automatisering BV, Harmelen, the Netherlands) were used to continuously monitor eating, rumination, and ear temperature in 37 crossbred Angus steers classified as High Growth (HG; 295 ± 34 kg) and 22 steers classified as Moderate Growth (MG; 274 ± 28 kg). Individual feed intake and feeding events were also recorded using Insentec automated feeding stations (Insentec BV, Marknesse, the Netherlands). Data from CowManager and Insentec were matched by animal, date, and hour and agreement between platforms was assessed using Pearson correlation coefficients and Bland–Altman plots. Predictive ability of CowManager eating and Insentec eating time for intake recorded by Insentec was evaluated with linear regression models. A repeated measures mixed model was used with phenotype, hour, and phenotype × hour interaction as fixed effects, and animal as the repeated measure. Significance was determined via t-tests and pairwise contrasts. Event duration measured by Insentec feeding stations was a strong predictor of dry matter intake (R2 = 0.76, P < 0.001). Eating time from CowManager sensors had a modest correlation with intake (R2 = 0.07, P < 0.001), rumination was a poor predictor (R2 = 0.00, P = 0.07), and combining rumination plus eating time from CowManager did not substantially improve prediction (R2 = 0.08, P < 0.001). Agreement between Insentec and CowManager eating time was moderate (Pearson’s r = 0.35, P < 0.05). Phenotypic comparisons indicated that high growth steers exhibited significantly higher ear temperature (P = 0.02), while main effects of phenotype on rumination (P = 0.36), intake (P = 0.94), and eating (P = 0.48) measured with CowManager were not significant in the repeated measures model. Significant diurnal (hourly) variation was observed for eating (P < 0.001), rumination (P < 0.001), and temperature (P < 0.001), with some significant phenotype-by-hour interactions (for rumination, P < 0.001, for temperature, P = 0.03). Overall, phenotype effects were modest, and the experiment reinforced the reliability of Insentec for intake prediction. Growth phenotype is associated with subtle differences in behavior and physiology, rather than large variations in feeding activity. Real-time feed intake monitoring technologies must be further developed and validated for use in grazing systems, as most current solutions are limited to enclosed or feedlot environments.