Genetic background of cow-human, cow-cow, and cow-technical system behavior responses.

The aims of the present study were to infer genetic parameters for behavior traits from the 3 behavior categories a) cow-human, b) cow-cow and c) cow-technical system interactions, and genetic associations between behavior pattern and milk production traits of dairy cows kept in herds with automatic milking system (AMS). Behavior and production data were recorded considering 3,647 Holstein (HOL) cows from 25 German AMS herds. Cow-human behavior traits included scores for avoidance distance (AVOIDIS), tolerance to tactile interaction (TTI) and behavior after released from the headlock (RELBH) from 2 recording dates. Cow-cow interaction traits for agonistic behavior patterns included scores for the same cow as actor (ACT) and as recipient (RECEIP). The cow-technical system behavior traits comprised visiting frequency in the milking robot (MILK-VIS), successful milkings (MILK-FREQ), AMS efficiency (MILK-EFF) reflecting milk yield per total box time and incomplete milkings (MILK-INCOMP), complemented with sensor behavior traits for rumination (RUM-TIME) and feeding duration (FEED-TIME). Production traits were daily milk yield (MY), fat percentage (Fat%) and protein percentage (Prot%) directly generated in the milking robot. Linear mixed models were applied to infer the impact of classical fixed effects of AMS-herd, season, parity and DIM on all behavior traits, indicating significantly lower least squares means for AVOIDIS in early parities and more problematic RELBH during summer. Genetic parameters were estimated by applying single-trait and multiple-trait repeatability models using a single-step approach simultaneously integrating pedigree and genomic 50K SNP data. Heritabilities for behavior traits from the cow-human and cow-technical system categories were low to moderate (0.03 to 0.35), indicating the potential for genetic improvement, especially for MILK-EFF (0.35 ± 0.031), FEED-TIME (0.30 ± 0.083) and RUM-TIME (0.26 ± 0.040). For cow-cow interaction traits, the heritability was 0.16 ± 0.031 for ACT, but close to zero for RECEIP. Phenotypic correlations among behavior traits were smaller than the respective genetic correlations. Strongest correlations were found between traits for the same behavior recorded in very dense intervals (e.g., genetic correlation of 0.81 ± 0.128 between avoidance distances at the same recording date), indicating consistency over repeated measurements. Across behavioral categories, genetic correlations were generally low to moderate, ranging from -0.26 ± 0.116 (RELBH with MILK-INCOMP) to 0.57 ± 0.125 (second TTI observation at the recording date with MILK-INCOMP). Based on their moderate heritabilities and mostly neutral genetic relationships with production traits, the cow-human interaction traits RELBH and TTI, and MILK-EFF from the cow-technical system behavior category, were identified as the most promising candidates for inclusion in AMS-specific breeding indices. In the present study, we estimated a broad portfolio of genetic covariances among cow behavior patterns from different behavior categories and with production traits, but for the construction of overall net merit indexes, genetic correlations with all other functional traits are imperative.