Genetic parameters of production traits in automatic and conventional milking systems

Milk recordings of 1297 animals from 23 farms with automatic (AMS) and 20555 animals from 616 farms with conventional milking systems (CMS) were analysed. The data was collected within the years 2001 – 2003. Recordings were available for milk-kg, fat and protein content from the first three test days of each cow after calving. Genetic parameters between AMS and CMS recordings for each single test day and with a fixed regression test day model were calculated Heritabilities of the three test days were in an expected range from .182 and .418. Genetic correlations were high with values near 1 for milk-kg and fat-%. Values for protein-% ranged between .760 and .881. Heritabilities for the fixed regression test day model with all three test days as repeated measures were for milk-kg .242 (AMS) and .201 (CMS), for fat-% .316 (AMS) and .254 (CMS) and for protein-% .202 (AMS) and .337 (CMS). Genetic correlations between AMS and CMS were 1.00 (milk-kg), .999 (fat-%) and .994 (protein-%). It can be concluded, that under this model no genetic difference can be found between AMS and CMS. Therefore genotype by environment interaction between automatic and conventional milk recording systems is irrelevant for milk yield and fat percentage and small for protein percentage. INTRODUCTION Dairy production in Europe is based on strict economic limits. Falling prices for milk force the farmers to invest in cost reducing technologies and to make the herd management more efficient. Milking robots used for automatic milking systems (AMS) eventually are an effective way to reduce the production costs by reducing the number of workers involved in the production system. These milking systems also have positive effects on the animals, by increasing the average number of milkings per day, which is a decision of the cow and no longer of man. The different numbers of milkings and other features of AMS raise the question, if AMS and conventional milking systems (CMS) can be directly compared. Especially for breeding value estimation, the new milking system needs to be tested for additional effects. Environmental effects are already included in the breeding value estimation, but if a performance from AMS is a different trait than from CMS, the whole breeding value estimation needs to be adapted for the new milking technologies. Most of the selection decisions are made on base of conventional milking systems. It has to be investigated, if the correlated response in AMSperformance is satisfying in comparison to direct selection on AMS. Differences between AMS and CMS performances are frequently reported. Phenotypic comparisons between AMS and CMS somatic cell counts (SCC) show a significant lower SCC in AMS systems (BERGLUND et al., 2002). In case of assuming different traits for AMS and CMS the milk quality can be different in the two systems and needs to be considered in the selection process. The aim of this study was to estimate genetic parameters of AMS and CMS milk recordings. MATERIAL AND METHODS Milk recordings of 1297 cows from all 23 farms with AMS and 20555 cows from 616 representative farms with CMS were analysed. The data were collected within the years 2001-2003. Recordings were available for milk-kg, fat and protein content and somatic cell count (SCC) from the first three test days of each cow after calving. SCC was used as log10 of SCC. Table 1 shows means and standard deviations of the three test days in each milking system. Table 1: Means and standard deviation of the milk recordings for each test day and milking system x (CMS) s (CMS) x (AMS) s (AMS) Test day 1 milk-kg 26.5 5.46 23.2 6.07 fat-% 4.29 0.79 4.20 0.78 protein-% 3.23 0.29 3.26 0.28 SCC 4.95 0.52 4.96 0.51 Test day 2 milk-kg 28.36 5.70 27.86 6.19 fat-% 3.88 0.59 3.84 0.61 protein-% 3.17 0.25 3.20 0.25 SCC 4.81 0.48 4.86 0.47 Test day 3 milk-kg 27.46 5.77 27.77 6.24 fat-% 3.90 0.62 3.87 0.64 protein-% 3.30 0.26 3.30 0.25 SCC 4.83 0.48 4.87 0.49 The three test days make two different statistical models possible. The first model treated the test day records as different traits. Genetic parameters for each test day record were calculated separately. The model included following effects: e a ci b ca b S H y k k k j i ijk + + ⋅ + ⋅ + + = 2 1 where: yijk = phenotypical performance of each animal Hi = fixed effect of the herd i Sj = fixed effect of the season j b1 = regression coefficient on age of first calving cak = age of first calving of cow k b2 = regression coefficient on calving interval cik = calving interval of cow k ak = random effect of the animal e = residual effect The analysis were calculated in bivariate runs, considering the same trait in AMS and CMS systems. The second model treated the three successive test days as repeated measurements of the same trait. The model was set up in analogy to the test day model with fixed regression lactation curves.