The number of dairy cows milked in automatic milking systems (AMS) is steadily increasing in Norway. Capacity and efficiency of AMS are highly dependent on the individual cow's milking efficiency, such as milking speed and occupation time in the milking robot. Cows meet new challenges in herds utilizing AMS. Consequently, new or revised traits may be needed for genetic evaluation of dairy cattle. The AMS records relevant information on an individual cow basis. The aims of this study were to estimate genetic parameters of new automatically recorded milkability and temperament traits. Data from 77 commercial herds with Norwegian Red dairy cattle were analyzed by mixed linear animal models. The final data set contained 1,012,912 daily records from 4,883 cows in first to ninth lactation. For variance component estimation, univariate and bivariate models were used. Daily records of box time (BT), average flow rate (FR), kilograms of milk per minute of box time (MEF), handling time (HT), log-transformed HT, milking frequency, and milking interval were analyzed with repeatability models. Among these traits, FR, BT, and MEF showed the highest heritabilities of 0.48, 0.27, and 0.22, respectively, whereas heritability of log-transformed HT, HT, milking frequency, and milking interval was low (0.02-0.07). Unsuccessful milkings expressed as rejected milkings, incomplete milkings (IM), milkings with kick-offs (KO), and teat not found also showed low heritabilities (0.002-0.06). Due to low frequency, KO, rejected milkings, IM, and teat not found were also analyzed as proportions per lactation, which resulted in slightly higher heritability estimates. Genetic correlations were favorable and intermediate to strong between BT, HT, MEF, and FR with absolute values above 0.50. Intermediate and favorable correlations were found for IM and KO with BT, HT, MEF, and FR. Cow milkability in AMS can be improved by selection for reduced number of unsuccessful milkings, faster FR, increased MEF, and shorter BT and HT. Our results confirm that automatically recorded data on milkability and temperament can be valuable sources of information for routine genetic evaluations and that milking efficiency in AMS can be genetically improved.

The aim was to investigate whether subjectively scored milking speed, temperament, and leakage are genetically the same trait when measured in different milking systems. Data were provided by the Norwegian Dairy Herd Recording System and included a total of 260,731 first-parity Norwegian Red cows calving between January 2009 and February 2019 and milked either in a traditional milking system (milking parlor or pipeline) or by an automatic milking system (AMS). Genetic parameters were estimated and lower heritabilities and less genetic variation were found for the 3 traits when measured in AMS herds. The heritability of temperament, leakage, and milking speed were 0.05, 0.04, and 0.22, respectively, with data from AMS herds; and 0.09, 0.14, and 0.27, respectively, with data from cows milked in traditional milking systems. The genetic correlations between temperament and leakage (−0.19), between milking speed and leakage (−0.88), and between milking speed and temperament (0.30) in AMS were slightly stronger than between the corresponding traits assessed in other milking systems (−0.15, −0.82, and 0.16, respectively). The genetic correlations between traits across milking systems were strong: 0.98, 0.96, and 0.86 for milking speed, leakage, and temperament, respectively. Strong correlations indicate that the traits were almost genetically similar despite being scored in different milking systems. The rank correlations among estimated sire breeding values were strong: 0.98 and 0.99 for milking speed and leakage, with little or no reranking of bull performance across milking systems. Temperament had the lowest genetic correlation (0.86) and rank correlation (0.91) across milking systems. These data suggest that AMS farmers evaluate temperament slightly differently from farmers using other milking systems or that different aspects of temperament are important for farmers with AMS.

The current study aimed to investigate new udder health traits based on data from automatic milking systems (AMS) for use in routine genetic evaluations. Data were from 77 commercial herds; out of these, 24 had equipment for measuring online cell count (OCC), whereas all had data on electrical conductivity (EC). A total of 4,714 Norwegian Red dairy cows and 2,363,928 milkings were included in the genetic analyses. Electrical conductivity was available on quarter level for each milking, whereas OCC was measured per milking. The AMS traits analyzed were log-transformed online cell count (lnOCC), maximum conductivity (ECmax), mean conductivity (ECmean), elevated mastitis risk (EMR), and log-transformed EMR (lnEMR). In addition, lactation mean somatic cell score (LSCS) was collected from the Norwegian dairy herd recording system. Elevated mastitis risk expresses the probability of a cow having mastitis and was calculated from smoothed lnOCC values according to individual trend and level of the OCC curve. The udder health traits from AMS were analyzed as repeated milkings from 30 to 320 DIM, and LSCS as repeated parities. In addition, both ECmax and lnOCC were analyzed as multiple traits by splitting the lactation into 5 periods. (Co)variance components were estimated from bivariate mixed linear animal models, and investigated traits showed genetic variation. Estimated heritabilities of ECmean, ECmax, and lnEMR were 0.35, 0.23, and 0.12, respectively, whereas EMR and lnOCC both showed heritabilities of 0.09. Heritability varied between periods of lactation, from 0.04 to 0.13 for lnOCC and from 0.12 to 0.27 for ECmax, although standard errors of certain periods were large. Genetic correlations among the AMS traits ranged from 0 to 0.99. The genetic correlations between EC-based traits and OCC-based traits in AMS were 0. Genetic correlations with LSCS were favorable, ranging from 0.37 to 0.80 (±0.11-0.22). The strongest correlation (0.80 ± 0.13) was found between LSCS and lnEMR. Results question the value of ECmax and ECmean as indicators of udder health in genetic evaluations and suggest OCC to be more valuable in this manner. This study demonstrates a potential of using AMS data as additional information on udder health for genetic evaluations, although further investigation is recommended before these traits can be implemented.