Methodology for quantifying the behavioral activity of dairy cows in freestall barns

The objectives of this study were to 1) evaluate the validity of automated monitoring systems as assessment method for the behavioral activity of dairy cows compared with video recording, and 2) determine the sampling intervals required to obtain reliable estimates of the daily behavior. To determine lying, standing, and walking, 12 cows were equipped with automatic recording devices (IceTag = 12 cows, HOBO Pendant G = 5 cows), and their behavior was simultaneously recorded using a video recording system. The correspondence between the IceTag, HOBO logger, and video recording data was analyzed using 2 x 2 contingency tables, and we determined the sensitivity, specificity, and predictive value (positive and negative). Both types of loggers demonstrated high sensitivity (Sen >= 0.961) and specificity (Sp >= 0.951) for lying and standing behaviors with predictive values near 1.00. The HOBO logger can accurately describe the laterality of lying behavior, whereas the IceTag device inadequately recorded walking, with probability predictive values = 0.978) was found between the total lying times from data on a per-second basis and estimates obtained by 1, 2, 3, 4, 5, 10, and 15 min sampling intervals. The sampling intervals of 1 and 2 min were comparable for all aspects of lying behavior (R-2 >= 0.813; P > 0.05 for slope = 1, intercept = 0). Long sampling intervals (30 and 60 min) showed positive relationship for estimating time spent lying and standing (R-2 >= 0.774), but were inappropriate for predicting these behaviors, because they lacked accuracy and precision. Both the IceTag and HOBO logger accurately measured all aspects of lying and standing behavior. Reliable estimates of lying and standing time can be generated using relatively short interval lengths (e.g., 3, 4, 5, 10, or 15 min). Shorter sampling intervals (