7 Assessment of a Chew-Sensor Technology for Quantifying Feeding Behaviors in Sheep

Abstract

Abstract Precision livestock management technologies for remote automated monitoring of feeding behavior can be utilized to improve animal health and production efficiency. The RumiWatch System (Itin+Hoch GmbH, Liestal, Switzerland) is a chew-sensor technology that has been validated for use in cattle and horses. The objective of this study was to validate the capability of the RumiWatch System to accurately quantify feeding behaviors in sheep. It was hypothesized that this chew-sensor technology would accurately report feeding and rumination time. Twelve Hampshire ewes (BW: 63 ± 2.0 kg; Age: 202 ±.1 d) were randomly assigned to four observational groups (n = 3 ewes∙group-1). Groups were pen-housed (4.8 x.5 m) and fed one concentrate meal (0.91 kg∙ewe-1∙d-1) in addition to ad libitum hay. One group per day was fitted with sheep-adapted prototype RumiWatch halters and observed over three 2-h periods using-min scan sampling, with each group observed for two days. Behaviors were classified as eating, ruminating, or other activity. Observations were recorded via Behavioral Observation Research Interactive Software (BORIS; v.7.13.8, Torino, Italy). Raw data collected using the prototype halters were converted using two versions of RumiWatch Converter software (v.7.3.2 and v.7.3.36). Agreement between visual observations and the RumiWatch output was evaluated in R (v. 4.0.1; R Foundation for Statistical Computing, Vienna, Austria). This included percent agreement and Cohen’s Kappa for-min behavior classifications and Pearson’s correlation coefficient (r) and concordance correlation coefficient (CCC) for hourly eating and rumination time. Agreement between visual observation and the two converter versions for time spent performing other behaviors was also assessed. Percent agreement was 81.4% and 86.9% for v.7.3.2 and v.7.3.36, respectively. Cohen’s Kappa for both versions indicated substantial accordance of observations and classified behaviors (v.7.3.2: κ = 0.71; v.7.3.36; κ = 0.79). Pearson correlations for eating time were r = 0.95 (v.7.3.2) and r = 0.96 (v.7.3.36; P < 0.0001). For rumination time, correlations were r = 0.88 (v.7.3.2) and r = 0.96 (v.73.36; P < 0.0001). The CCC between observations and system-recorded eating time indicated very high agreement regardless of converter version (v.7.3.2: CCC = 0.92; v.7.3.36: CCC = 0.96). High or very agreement was also found for rumination time (v.7.3.2: CCC = 0.73; v.7.3.36: CCC = 0.92). Agreement was high for observations and time assigned as other activity regardless of converter version (v.7.3.2: r = 0.96 [P < 0.0001], CCC = 0.90; v.7.3.36: r = 0.98 [P < 0.0001], CCC = 0.94). Results of this study demonstrate that the RumiWatch System can accurately quantify time spent feeding and ruminating in sheep, with future applications for both researchers and producers.