Validating the accuracy of activity and rumination monitor data from dairy cows housed in a pasture-based automatic milking system.

Behavioral observations are important in detecting illness, injury, and reproductive status as well as performance of normal behaviors. However, conducting live observations in extensive systems, such as pasture-based dairies, can be difficult and time consuming. Activity monitors, such as those created for use with automatic milking systems (AMS), have been developed to automatically and remotely collect individual behavioral data. Each cow wears a collar transponder for identification by the AMS, which can collect data on individual activity and rumination. The first aim of this study was to examine whether cow activity levels as reported by the AMS activity monitor (ACT) are accurate compared with live observations and previously validated pedometers [IceQube (IQ), IceRobotics, Edinburgh, UK]. The second aim of the study was to determine if the AMS rumination monitors (RUM) provide an accurate account of time spent ruminating compared with live observations. Fifteen lactating Holstein cows with pasture access were fitted with ACT, RUM, and IQ. Continuous focal observations (0600-2000 h) generated data on lying and active behaviors (standing and walking), as well as rumination. Activity recorded by live observation and IQ included walking and standing, whereas IQ steps measured cow movement (i.e., acceleration). Active behaviors were analyzed separately and in combination to ascertain exactly what behavioral components contributed to calculation of ACT "activity." Pearson correlations (rp) were computed between variables related to ACT, RUM, IQ, and live observations of behavior. A linear model was used to assess significance differences in the correlation coefficients of the 4 most relevant groups of variables. Significant but moderate correlations were found between ACT and observations of walking (r(p)=0.61), standing (r(p)=0.46), lying (r(p)=-0.57), and activity (r(p)=0.52), and between ACT and IQ steps (r(p)=0.75) and activity (r(p)=0.58) as well as between RUM and observations of rumination (rp=0.65). These data indicate that ACT and RUM do reflect cow walking and rumination, respectively, but not with a high degree of accuracy, and lying cannot be distinguished from standing.