Sire evaluations at different levels of milk production

Data were 49,693 first 305-d lactation milk kg, fat kg and fat percentage from the Gdańsk region. Black and White cows were daughters of 523 Black and White and 27 Holstein-Friesian bulls. Herds were stratified into three levels, based on the mean milk yield. Genetic and residual variances were estimated by the restricted maximum likelihood method (REML). The model included effects of herd-year-season of calving, and of genetic groups. The same model was used for BLUP evaluation of sires. AU calculations were done on linear and logarithmic scales, within the three production levels. Heritabilities of milk yield, fat yield and fat content increased with increasing mean production. Similar estimates were obtained on the linear and log scales. Genetic variance increased on both scales, the same way heritabilities did. Error variances of milk and fat yield followed the same pattern on the linear scalę. On the log scalę for those two traits, the highest error variance was obtained at the medium production level. Correlations between sire breeding values estimated at different production levcls were close to expected correlations. Genetic correlations between levels were Iow for fat yield, especially between the high and the other levels. This seems to suggest a certain amount of genotype-environment interaction. Evaluation and selection of sires at Iow production might adversely affect genetic progress.


INTRODUCTION
Sire evaluation for production traits in Poland is based on first 305-d lactations of daughters from large state farms.The cow population is characterized by Iow average overall yields and relatively large variation among herds.In several studies it was shown that the standard deviation inereases with the inerease in mean milk and fat yield.Also genetic components of variance were often higher at higher production levels (Danell, 1982;Hill et al., 1983;Lofgren et al., 1985;Mirandę and Van Vleck, 1985;De Veer and Van Vleck, 1987).Majority of studies concerning expression of a genotype in different environments and heterogeneity of variances were carried out on much higher production than that studied in this paper.
The purpose of this study was to determine the effects of different production levels on genetic and residual variance components and BLUP sire evaluation results in a population with Iow overall milk yield.

DATA AND METHODS
Data were 49,693 first 305-d lactation milk kg, fat kg and fat percentage from the Gdańsk region.Black and White cows were daughters of 523 Black and White and 27 Holstein-Friesian bulls.Genetic groups were defined according to the strain and bulls' year of birth.Herds were stratified into three levels, based on the mean milk yield: below 3,000 kg (488 bulls with 16,464 daughters; 3,001 to 4,000 kg (544 bulls with 24,137 daughters) and above 4,001 kg (331 bulls with 9,092 daughters).
Genetic and residual variances were estimated by the restricted maximum likelihood method (REML).The model included effects of herd-year-season of calving, and of genetic groups.The same model was used for BLUP evaluation of sires.Ali calculations were carried out on linear nad logarithmic scales, within the three production levels.
Genetic correlations between production levels were approximated by the adjustment procedurę given by Cało et al. (1973).Expected correlations were calculated as a function of the accuracy of sire transmitting abilities (Hickman et al., 1969).

RESULTS AND DISCUSSION
Average milk production at three levels (Iow, medium and high) and numbers of bulls and daughters are given in Table 1.This data was used to evaluate the genetic parameters.Table 2 shows the estimates of sire and error variances and heritabilities by production level.Heritability of milk yield increased with the production level, from 0.18 to 0.38.Both components of variance also increased, with the genetic component growing faster.A similar pattern was found for fat yield but heritabilities were lower (from 0.09 to 0.23).Heritabilities of fat content for two lower production levels were smali (0.11 and 0.15).Only in the highest production level heritability reached the magnitude usually reported in the literaturę (0.41).
Logarithmic transformation does not change heritabilities of studied traits.The changes occured in the relative magnitude of variance components.In milk and fat yield the error variance on the log scalę was highest at the medium level, lowest at the Iow level and intermediate at the high level.Fat content had the highest error variance at the Iow level and slightly lower and of the same magnitude at the medium and high levels.Sire variance showed the same increasing pattern on the log scalę and on the linear scalę for all traits.
The increase of heritabilities with increasing production level found in our study is in agreement with other research (see e.g.Dannel, 1982;Mirandę and Van Vlec, 1985;De Veer and Van Vleck, 1987).In these papers it was also shown that sire variance was related to the mean.However, for records transformed to logarithms, the error variance usually decreased as production level increased (Mirandę and Van Vleck, 1985) or remained on the same level (De Veer and Van Vleck, 1987).In our study the error variance for milk and fat yield was smaller for the high than for the medium level.The Iow level had the smallest error variance.This seems to show that the multiplicative model is not ąuite adeąuate to describe the records.
Genetic and phenotypic correlations among three traits within each produc tion level showed similar magnitude.
Table 3 presents the correlations between BLUP predicted differences of the same bulls evaluated on records of daughters split according to the production level.The values of all correlations were close to their expected values, on both scales.The genetic correlations between levels of production were near unity for milk yield.However, for fat yield correlations were much lower.The highest genetic correlation on the untransformed records was found between the Iow and medium levels (0.73), correlations between high and medium and high and Iow were much less than unity (0.68 and 0.66 respectively).In other studies the correlations among sire values at different production levels were usually above 0.8 (Danell, 1982), and their authors concluded that genotype-environment interaction does not have to be considered in the selection programs.In our data this is the case for milk yield, but much lower genetic correlation coefficients for fat yield showed that ranking of bulls at the high level was substantially different from that at the Iow and medium level.This seems to suggest that a certain amount of interaction in fat yield is present in our data and is connected with the lower production levels.Evaluation and selection of sires at Iow production might adversely affect genetic progress.

CONCLUSIONS
Heritabilities of milk yield, fat yield and fat content increased with increasing mean production.Similar estimates were obtained on the linear and log scales.
Genetic variance increased on both scales in the same manner as heritabilities did.Error variances of milk and fat yield followed the same pattern on the linear scalę.On the log scalę for those two traits, the highest error variance was obtained at medium production.
Correlations between sire breeding values estimated at different production levels were close to expected correlations.

TABLE 3
Correlations between buli BLUP breeding values (r s ), genetic correlations (r g ) and expected correlations (r exp ) estimated at different production levels (on linear and long scalę) with minimum 10 daughters/bull