Automatic milking unit preference of Holstein, Jersey, and Holstein-Jersey crossbred cows in a batch milking system

The efficiency of automatic milking systems depends on the continuous flow of cows, which may be affected by specific cow conducts, including milking unit preference. Moreover, providing resources that favor the expression of natural cow behaviors during the milking process may result in improved animal welfare. The objective of this study was to analyze the selection behavior for automatic milking units, comparing preference consistency across the lactation of 3 genetic groups (Holstein [HO], Jersey [JE], and Holstein × Jersey [HJ]) in a multibreed organic dairy farm using a batch milking system. To expand the main objective, a secondary aim was to examine the relationship between milking unit selection behavior and premilking waiting time (WT). The study included data from 1,762,461 milking events in 1,355, 1,876, and 475 HO, JE, and HJ cows, respectively, from May 2023 to September 2024 in a commercial organic grass-fed dairy in Texas in the United States. Cows were moved to the waiting area of the milk center twice per day, where they could select their milking visits among 22 automatic milking units. Cow preferences were analyzed at 3 levels: automatic milking unit (n = 22), barn location (east, n = 11; west), and robotic arm configuration (left, n = 11; right). Milking visit information was collected from the management software to establish the frequency of specific milking unit usage per cow during the study period. Subsequently, the frequencies for selection of the top choice for 1, 3, and 5 automatic milking units, the top choice for barn location, and the top choice for robotic arm configuration were calculated for each cow. Preference consistency scores (PCS) were calculated as a ratio between the excess frequencies of the first choice over the base frequency of any option other than the first choice. The PCS calculations considered the frequency of access to each automatic milking unit, barn location, and robotic arm configuration in 30-d periods until 305 DIM. Premilking waiting times were calculated for each cow as the difference between the entry time into the waiting area of the milk center, indicated by a pedometer affixed in the rear leg of each cow, and the start of the corresponding milking, as indicated by the milking unit in use. Least squares means for PCS were calculated within parity category by genetic group for each 30-d period and compared using repeated-measures ANOVA. Subsequently, PCS and WT values were categorized into quartiles, and LSM for PCS and WT were calculated for each of the PCS by WT combinations within each parity category. As indicated by the chi-squared test of independence and the ANOVA, the frequency of selection for the top choice and the PCS were different from those expected by random selection of automatic milking unit, barn location, and robotic arm configuration, signifying different levels of selection behavior in the study cows. Overall, HO cows evidenced the greatest PCS values, indicating more consistent behaviors in milking unit preference. Additionally, the analysis of the association between PCS and WT indicated that greater PCS were associated with shorter WT. The resulting PCS suggest variable degrees of consistency in the selection behavior for automatic milking units in dairy cows. The selection behaviors identified in this study may have implications for cow welfare, as well as for improving cow traffic and system efficiency.