Lameness changes the behavior of dairy cows: daily rank order of lying and feeding behavior decreases with increasing number of lameness indicators present in cow locomotion

Abstract

Lameness is a major animal welfare problem in modern dairy farms. Although association between cow behavior and lameness is proven, functioning and reliable on-farm applications for automatic lameness detection is still lacking because large within and between farm variation in cow behavior exists. There were two aims in this study: 1) to investigate if a simple ranking of cows in an order based on their time spent on a specific behavior or number of behavior bouts could be used for normalizing behavior data, and 2) to study how daily lying and feeding behavior, and their rank order variables are affected by lameness. We followed in total 84 cows divided in two half-year trials (N = 45 and N = 49, ten cows enrolled in both trials). Cows were locomotion scored fortnightly and their lying,roughage feeding behavior, and step count were measured continuously with automatic sensors. In addition to using absolute values of daily behavior, animals were ranked in ascending order within experimental cows in the group for each day based on their absolute value (hours/number of bouts/number of steps) of the behavior in question. Daily behavioral and cow level factors were merged with temporally closest locomotion score. Two sets of linear mixed models were fitted: 1) to investigate the effect of trial on absolute and rank order values of behavior, and 2) to investigate effect of locomotion score, parity, breed and lactation stage on absolute and rank order values of behavior. There was a significant difference between the trials in lying behavior of the cows, but not in feeding behavior; cows spent more time lying down in the first trial (P= 0.0027). This difference was not evident in rank order values. Increasing the number of lameness indicators was associated with increasing lying time, increase in number of lying bouts, decreasing roughage feeding time and decrease in feeding bouts (P<0.0001). Similar differences were observed in rank order values of behaviors (P<0.0001). No consistent change in step count was observed with respect to increasing number of lameness indicators. Based on these data, ranking the cows within the herd based on their behavior did normalize the data between two trials, and rank order values behaved similarly to absolute daily behavior with respect to a cow's lameness status. Thus, using rank order variables could address some issues in development of automated lameness detection systems arising from within and between herd variability in dairy cow behavior.