Livestock Research for Rural Development 16 (4) 2004

Citation of this paper

Variance heterogeneity for milk yield in Brazilian and Colombian Holstein herds  

M F Cerón-Muñoz, H Tonhati*, C N Costa**, D Rojas-Sarmiento*** and C Solarte Portilla****

Facultad de Ciencias Agrarias - Universidad de Antioquia- Colombia.\
mceronm@universia.net.co
*Dept. Zoot , FCAV/UNESP, 14884-900 Jaboticabal-SP, Brazil.
Embrapa Gado de Leite, 36038-330 Juiz de Fora-MG, Brazil.
***Asociación Holstein de Colombia
****Facultad de Ciencias Pecuarias - Universidad de Nariño
(Part of doctoral Thesis of 1st author - FCAV-Unesp. Supported by FAPESP - Brazil.)


Abstract

Heterogeneity of variances for milk yield (MY) was determined for Brazilian and Colombian herds. The herds were grouped as high or low variability within each country, using as criterion the phenotypic standard deviation (PSD) of MY in the contemporary groups of cows, from the first to the sixth calving. Brazilian and Colombian herds with PSD greater than 1,168 kg or 1,012 kg, respectively, were classified as high variability while the herd groups with values lower than those were classified as low variability. The genetic parameters for MY within each herd group were estimated using bivariate analysis in an animal model and the restricted maximum likelihood method with a derivative free algorithm, using 72,280 first lactations of cows, daughters of 1,880 sires.

Heterogeneous variances were found, and Brazilian herds with high PSD had the greatest additive and residual genetic variances and heritability coefficients for MY. MY genetic correlation coefficients between herds of high and low variability within each country were 0.96 and 0.93 while between countries they varied from 0.72 to 0.81, suggesting that there was a reclassification of animals in the two countries and also heterogeneity of variances. This phenomenon leads to the questioning of the strategy of imported semen usage and the need to do genetic evaluations to identify sires with greater genetic potential for (sub) tropical environmental conditions.

Key words: Dairy cattle, genotype - environment interaction

Introduction

Animal genetic improvement programs involve comparisons among individuals of contemporary groups maintained in different environments, which could differ on averages as well as on variability among them; therefore, the animals could be erroneously classified, since the sires with most of their daughters in herds with greater variances would tend to have their genetic attributes overestimated (Hill 1984; Vinson 1987).

For Van Vleck (1988), genotypical and phenotypical variances are, in most cases, different from farm to farm. In studies referring to herd grouping by phenotypical standard error average of the contemporary groups (herd-year) done by Stanton (1990) in Colombia, Mexico, Puerto Rico and United States of America, by Costa (1999) in Brazil and USA, and by Cienfuegos-Rivas et al (1999) in Mexico and USA, it was concluded that the response differences to the use of sires in (sub) tropical countries seem to be associated with heterogeneous variances of the herds. This brings as a result wide phenotypical differences among the daughters. This problem becomes worse as the environment variation range increases where the improved genetic material would be used (Stanton 1990). Therefore, by ignoring variance heterogeneity, there will be consequences in resulting selection and genetic gain, reducing the effectiveness of the genetic breeding program (Vinson 1987; Van Vleck 1987).

In a study by Stanton (1990), the phenotypical responses of the daughters of different grouping classes in Latin American herds were smaller than those in American herds, suggesting that the genetic variances found in (sub) tropical environments are smaller than in temperate environments, since Holstein cattle under restricted management conditions did not present adequate physiological development and do not have the opportunity to express their production genetic potential. This phenomenon reduces population variability (McDowell et al 1976; Lopes-Villalobos et al 2001; Holmann et al 1990; Costa 1998).

Lopes-Villalobos et al (2001) and Holmann et al (1990) evaluated the profit of investing in North American semen for use in in (sub) tropical countries and concluded that these importations should be used strategically in environments with excellent feed and management, in order to obtain a positive response on the selection of imported genetic material.

This study had as objectives to determine the heterogeneity of genetic and residual variances in Brazilian and Colombian herds and to estimate the genetic correlation of milk yield between high and low variability herds.


Materials and Methods

The study was done based on 180,522 records of milk yield of 94,558 Holstein cows from 775 Brazilian and 317 Colombian herds, derived from the data bases of the State Associations affiliated to the Brazilian Holstein Cattle Ranchers Association (Associação Brasileira de Criadores de Bovinos da Raça Holandesa (ABCBRH)) and made available by the National Animal Sciences Archives (Arquivo Zootécnico Nacional), managed by Embrapa Gado de Leite, Juiz de Fora, Minas Gerais, and from the archives of the Colombian Holstein Association (Asociación Holstein de Colombia (Asoholstein de Colombia)), from parities between 1980 and 1997. In Brazilian and Colombian Holstein dairy farms the cattle are mostly fed pasture and concentrates. Milk is produced on large and very intensive farms as well as in family systems.

Milk yield was adjusted for 305 day lactations, age, calving order and season, using correction factors suggested by Torres (1998) and Torres et al (2000a) for Brazil and by Cerón-Muñoz et al (2003) for Colombia.

The herds were grouped in high or low variability classes within each country, using as criterion the average phenotypical standard deviation (PSD) of milk yield for herd-year (HY) contemporary groups. For that purpose, PSD limits of 1,168 kg and 1,012 kg in Brazilian and Colombian herds, respectively, were considered. Herds with PSD values greater than those above mentioned were classified in the high PSD, while those with lower values were put in the low PSD group. The choice of PSD limits considered the condition of establishing in each group a similar number of sires, daughters and herds. Also, a minimum of 10 lactations of cows from the first to the sixth calving, by contemporary group, were considered.

After classifying the herds, only the information on the first lactation was used for genetic parameter estimation, discarding the contemporary groups with less than three records and sires with less than three daughters in by less than three herds. The numbers of herds were 307 and 468 in the high and low Brazil PSD and 139 and 178 in then high and low Colombian PSD, respectively.

The sires were grouped for origin and period of birth within origin (Table 1). Most of the Canadian and Americans sires had daughters in the two countries. The Brazilian and Colombian sires only had daughters in Brazil and Colombia, respectively; which, however, were descended from common ancestors (Canadian or American sires). The cows were grouped as: Holstein purebred or grade (genetic composition equal or less than 31/32 Holstein.

Table 1. Number of sires (S) and daughters (D) by sire genetic group in high and low variability herds from Brazil and Colombia

 

Genetic group

Brazil

 

Colombia

 

Common Sires

 

High Variability (HB)

 

Low  Variability (LB)

 

High Variability (HC)

 

Low Variability (LC)

 

HB-LB

HB-HC

HB-LC

LB-HC

LB-LC

HC-LC

S

D

S

D

S

D

S

D

Br (≤79).....

73

838

 

50

1837

 

-

-

 

-

-

 

40

0

0

0

0

-

Br (80-84)..

168

1922

 

136

2909

 

-

-

 

-

-

 

94

0

0

0

0

-

Br (85-93)..

157

2007

 

189

1641

 

-

-

 

-

-

 

83

0

0

0

0

-

Ca (≤76)....

65

1057

 

61

1598

 

40

579

 

35

620

 

57

25

21

23

20

32

Ca (77-81).

32

1744

 

32

1326

 

33

880

 

34

1013

 

31

14

14

15

15

33

Ca (82-93).

51

1696

 

59

641

 

51

889

 

59

699

 

47

26

31

29

38

45

Co (≤79)....

-

-

 

-

-

 

44

448

 

39

445

 

-

0

0

0

0

31

Co (80-84).

-

-

 

-

-

 

66

1090

 

52

817

 

-

0

0

0

0

41

Co (85-93).

-

-

 

-

-

 

63

803

 

63

814

 

-

0

0

0

0

36

Eu (≤72)....

87

1582

 

84

2260

 

75

781

 

65

920

 

71

40

33

38

35

49

Eu (73-75).

87

1364

 

84

2274

 

87

1160

 

88

1596

 

78

51

49

50

49

72

Eu (76-82).

236

8196

 

251

7134

 

231

3575

 

230

3398

 

222

134

132

142

138

197

Eu (83-93).

139

4491

 

172

1698

 

140

2053

 

146

1485

 

132

62

58

80

77

112

Total  sires

1095

24897

 

1118

21058

 

830

11098

 

811

11807

 

 

 

 

 

 

 

Br= Brazilian sires; Ca= Canadian sires; Co= Colombian sires; Eu= North American sires.
HB = High variability Brazilian herds, LB = Low variability Brazilian herds, HC = High variability Colombian herds, LC = Low variability Colombian herds

 

The genetic parameters for milk yield in each herd group were estimated using bivariate analysis in an animal model and the restricted maximum likelihood method with an algorithm free of derivates. The fixed effects of HY, sire genetic group, and cow genetic group, and the random effects of animal and error were included. The model in a matrix notation for bivariate analyses of milk yield between Brazilian and Colombian herds of high and low PSD, can be represented as:

with

 

considering that:

where  is the additive genetic variance for characteristic i=1,2, and  is the additive genetic covariance between characteristics 1 and 2.

A is the relationship matrix and  is the direct product; and

where  is the residual variance for characteristic i=1,2.

The components of (Co)variance were estimated using the computer program MTDFREML (Boldman et al 1993).


Results and Discussion

Herds in the high variability group presented greater milk yield averages (Table 2). These results confirm that there is a positive relationship between the average and the phenotypical variance of the herds, as discussed by Hill et al (1983), Brotherstone and Hill (1986) and Boldman and Freeman (1988). Other studies did not highlight this relationship (Lofgren et al 1985).

In Brazil, genetic and residual variances of milk yield for herds in the group of low PSD were, respectively, 41% and 58% of the variances estimated for the herds in the group of high PSD. For low PSD in Colombia the respective values were 56 and 55% (Table 2).

Table 2. Estimates of (co) variance, heritabilities1 and genetic correlations for milk yield in Holstein cattle according to the herd groups of high and low phenotipical standard-deviation in contemporary groups in Brazil and Colombia.

 

Brazil

 

Colombia

High

Low

 

High

Low

Milk yield average, kg

6,932.47

6,174.69

 

6,416.23

5,196.63

Milk yield standard deviation, kg.

1,725.38

1,543.56

 

1,659.93

1.276.28

Coefficient of variation, %.

24.89

25.00

 

25.87

24.56

Additive genetic variance, kg2.

678,195.25

278,279.23

 

320,302.25

180,602.68

Residual variance, kg2

1,166,513.62

674,952.47

 

865,650.51

473,228.75

Heritability

0.37

0.29

 

0.27

0.28

 

Genetic correlations

 

Brazil

 

Colombia

Low

 

High

Low

Brazilian herds with high standard deviation

0.96

 

0.81

0.73

Brazilian herds with low standard deviations

 

 

0.77

0.72

Colombian hers with high standard deviation

 

 

 

0.93

1 The variances and heritability estimations are the means of the bivariate analyses

For Costa (1999), Stanton (1990) and Cienfuegos-Rivas et al (1999), differences in herd size and management practices have been associated with heterogeneous residual variances in Latin American herds. Smaller genetic variance in low PSD herds possibly could be due to the fact that the animals did not present the desired physiological development and did not have the opportunity of expressing their genetic production potential (McDowell et al 1976; Lopes-Villalobos et al 2001; Holmann et al 1990; Costa 1998).

However, in PSD herds of Latin American countries, animal performance is close to that of countries of origin, since the genetic potential is expressed and, consequently, the use of imported genetic material presents positive and expected results (Costa,1999; Stanton 1990; Cienfuegos-Rivas et al 1999). Greater variances in herds of high PSD than in low ones were also observed in exporting countries, indicating that in cattle populations, independently of country or genetic breeding program, the genetic potential is better expressed when environmental and management restrictions are less severe (Van Vleck 1988; Boldman and Freeman 1990; Sullivan and Schaeffer 1989; Hill et al 1983; Powell et al 1983; Dong and Mao 1990).

Greater heritability estimates were found in high PSD herds than in low ones in Brazil (0.37 and 0.29, respectively). A similar relationship of heritability estimates between groups classified by PSD was found by Torres et al (2000b), who used three herd classes (high, median and low PSD), finding heritability estimates of 0.30, 0.35, and 0.27, respectively and differed from the results found by Costa (1999), where herds of high PSD presented smaller heritability estimate (h2=0.22) than those of low variability (h2=0.30). In general, these differences can be caused by the distinct models used and the number of records considered.

In Colombia, the heritability was similar in the high and low PSD (0.27 and 0.28, respectively).

The genetic and residual variance were smaller in low PSD herds in both countries; however, as the herd variability increases, the genetic variance increases in the same proportion as the residual variance, which confirms that smaller variances are not necessarily associated to smaller heritability estimates for milk yield (Stanton 1990).

The observation of variance heterogeneity and/or different heritabilities in the high and low PSD, indicates that when ignoring the existence of heterogeneous variances, unexpected results are achieved by national genetic breeding programs (Hill 1984; Vinson 1987).

The genetic correlation coefficients for milk yield between high and low variability herds were 0.96 and 0.93 in Brazil and Colombia, respectively (Table 2). Although the genetic correlation estimates were close to unity, they were sufficiently low to significantly change the classification of sires between environments. Additionally, the existence of heterogeneous variances affects the performance magnitude of sires and cows in the environments where they are used (Costa 1999; Costa et al 2001).

The genetic correlation coefficients for milk yield between the high and low PSD groups of the two countries varied from 0.72 to 0.81.  According to Falconer (1952) and Dickerson (1962), the genetic correlations minor to one, suggest that, possibly, there was reclassification of animals. In this investigation, the Brazilian and Colombian Holstein cattle populations, presented reclassification, potentially reducing the benefits of the global semen importation strategy for these countries. In this context, national genetic breeding programs should be aware of the differences among environments, since there would be an efficiency reduction of sire and cow selection by assuming homogeneous variances when they are evidently heterogeneous (Garrick and Van Vleck 1987).

Greater variances in high PSD herds were found in the two countries, and this phenomenon was also observed in other countries registered as importers of genetic material, which leads to the questioning about the best strategy for use of imported semen, which is evaluated only in environments favorable for yield. There is a present need to do genetic evaluation to identify sires of greater genetic value in the environmental condition of the (sub) tropical countries and, in these evaluations, to consider the marked differences in production systems within each country. As a contrast, there is an opportunity to develop genetic breeding programs at the national and regional levels that would benefit the low PSD herds, since the genetic value of the sires would not be masked in the conditions of unfavorable production environments.


Conclusions


References

Boldman K G and Freeman A E 1988 Estimates of genetic and environmental variances of first and later lactations at different production levels. In: Animal Model Workshop, 1988, Edmonton, Proceedings p. 81.

Boldman K G and Freeman A E 1990 Adjustment for heterogeneity of variances by herd production level in dairy cow and sire evaluation. Journal of Dairy Science, Champaign, v.73, n. 2, p.503-512.

Boldman K G, Kriese L A and Van Vleck L D 1993 A manual for the use of MTDFREML: a set of programs of variances and co variances. Lincoln: Department of Agriculture Research Service. 120 p.

Brotherstone S and Hill W G 1986Heterogeneity of variances amongst herds for milk production. Animal Production, Edinburgh, v.48, n.3, p.283-291.

Cerón-Muñoz M F, Tonhati H and Costa C 2003 Factores de ajuste para producción de leche en bovinos Holstein Colombiano. Revista Colombiana de Ciencias Pecuarias, Medellín, v.16, n.1, p.26-32.

Cienfuegos-Rivas E G, Oltenacu P A, Blake R W, Schwager S J, Castillo-Juarez H and RuizF J 1999 Interaction between milk yield of Holstein cows in Mexico and the United States. Journal of Dairy Science, Champaign, v.82, n.10, p.2218.

Costa C N 1998 Genetic relationships for milk and fat yields between Brazilian and United States Holstein cattle populations. 1998. 175f. Thesis (Doctor of Philosophy)-Department of Animal Sciences, Cornell University, Ithaca.

Costa C N 1999 An investigation into heterogeneity of variance for milk and fat yields of Holstein cows in Brazilian herd environments. Genetics and Molecular Biology, Ribeirão Preto, v.22, n.3, p.375-381.

Costa C N, Castro R P and Haquihara K 2001 Genetic progress in Holstein cattle in Brazil. In: Congresso Holstein de las Américas, 6.,2001, São Paulo. Anais. p.106-120.

Dickerson G E 1962 Implications of genetic-environmental interactions in animal breeding. Animal Production, Edinburgh , 4:47-63.

Dong M C and Mao I L 1990Heterogeneity of (co)variance and heritability in different levels of intraherd milk production variance and of herd average. Journal of Dairy Science, Champaign, v.73, n.3, p.843.

Falconer D 1952 The problem of environment and selection. Amer. Nat., Chicago, 86:293-298.

Garrick D J and Van Vleck L D 1987 Aspects of selection for performance in several environments with heterogeneous variance. Journal of Dairy Science, Champaign, v.65, n.2, p.400-421.

Hill W G 1984 On selection among groups with heterogeneous variance. Animal Production, Edinburgh, v.39, n.3, p.473.

Hill W G, Edwards M R and Thompson R 1983Heritability of milk yield and composition at different levels and variability of production. Animal Production, Edinburgh, v.36, n.1, p.59-68.

Holmann F J, Blake R W and Milligan RA 1990 Economic returns from United States artificial insemination sires in Holstein herds in Colombia, Mexico and Venezuela. Journal of Dairy Science, Champaign, v.73, n.8, p.2179-2189.

Lofgren D L, Vinson W E and Pearson R E 1985Heritability of milk yield at different herd means and variances for production. Journal of Dairy Science, Champaign, v.68, n.10, p.2737-2739.

Lopes-Villalobos N, Garrick D J and Holmes C N 2001 Effects of importing semen of Holstein Friesian and Jersey bulls on the future profitability of an Argentine farm. Archivos de Zootecnia., Córdoba, v.50, n.191, p.311-322.

McDowell R E, Wiggans G W and Camoens J R 1976 Sire comparisons for Holstein in Mexico versus the United States and Canada. Journal of Dairy Science, Champaign, v.59, n.2, p.298-304.

Powell R L, Norman H D and Weiland B T 1983 Cow evaluation at different milk yields of herds. Journal of Dairy Science, Champaign, v.66, n.1, p.148-154.

Stanton T L 1990 Investigation of genotype by environment interaction for Holstein milk yield in Colombia, Mexico and Puerto Rico. 1990, 207f. Thesis (Doctor of Philosophy) - Department of Animal Sciences, Cornell University, Ithaca.

Sullivan P G and Schaeffer L R 1989 Regional heterogeneity of variances and its effect on Canadian Holstein sire evaluation. Canadian Journal of Animal Science, Ottawa. v.69, n.3, p.605-612.

Torres R A 1998 Efeito da heterogeneidade de variância na avaliação genética de bovinos da raça holandesa no Brasil. 1998. 133f. Tese (Doutorado em Ciência Animal) - Universidade Federal de Minas Gerais, Belo Horizonte.

Torres R A, Araújo C U e Costa C N 2000a Ajustamento da produção de leite para os efeitos simultâneos de ordem, idade e estação de parto. Revista Brasileira de Zootecnia, Viçosa, v.29, n.8, p. 2253-2259.

Torres R A, Torres Filho R, Vieira de Araújo C 2000b Heterogeneidade de variância e avaliação genética de bovinos da raça holandesa no Brasil. Revista Brasileira de Zootecnia, Viçosa, v.29, n.4, p. 1050-1059.

Van Vleck L D 1987 Selection when traits have different genetic and phenotypic variances in different environment. Journal of Dairy Science, Champaign, v.70, n.2, p.337-344.

Van Vleck L D 1988 Alternatives for evaluation with heterogeneous genetic and environmental variances. In: Animal Model Workshop. 1988, Edmonton. Proceedings p. 83.

Vinson W E 1987 Potential bias in genetic evaluations from differences in variation within herds. Journal of Dairy Science, Champaign, v.70, n. 11, p.2450.


Received 21 January 2004; Accepted 1 March 2004

Go to top