Heterogeneity of variance components for production traits in large‐scale dairy farms

The aim of this study was to estimate genetic parameters applying fixed regression and random regression test day models for protein yield using subsets stratified by herd size, average production level of herds in test day protein yield, and average age at first calving within herds. The data consisted of 64,394 Holstein cows located in one region in Eastern Germany including 690,553 first lactation test day production records. Estimates of variance components revealed heterogeneity of genetic variances across subsets and the highest additive genetic variances and heritabilities were found in larger herds with high production level and low age at first calving. Special cooperator herds for dairy cattle progeny testing programs will be a common practice in the near future. Suitable selection criteria for cooperator herds should consider genetic parameters, e.g. genetic variances and heritabilities within herds or subsets, to identify genetically superior sires as accurately as possible. As shown in the present study, the two‐step cluster analysis is an appropriate method to combine all desirable effects of herd parameters for a final selection of cooperator herds. One cluster included 44 large‐scale dairy farms with an average of 638 cows per farm and was also characterised by the highest protein yield, the lowest age at first calving, the highest heritability, and the highest additive genetic variance for test day protein yield within herds. A cluster analysis can be implemented in the work routine of dairy cattle breeding programs and delivers results in acceptable time. Rank correlations of estimated breeding values of bulls in different subsets as well as in different clusters were greater than 0.90, and therefore disprove any concerns for possible genotype by environment interaction between the test and the production environment. The permanent environment variance was slightly increased in strata indicating a better environment, whereas the residual variance component was similar for all subsets. There were no remarkable differences in genetic correlations between individual test days and test days 25 and 150, when analysing different subsets.