|Title||Selection for longevity in dairy cattle|
|Source||Agricultural University. Promotor(en): E.W. Brascamp; A.F. Groen. - S.l. : Vollema - ISBN 9789054858782 - 155|
Animal Breeding and Genomics
|Publication type||Dissertation, internally prepared|
|Keyword(s)||melkvee - selectief fokken - kunstmatige selectie - fokwaarde - gebruiksduur - productieve levensduur - kenmerken - heritability - dairy cattle - selective breeding - artificial selection - breeding value - longevity - productive life - traits - heritability|
This thesis deals with several aspects of longevity of dairy cattle. When breeding organizations want to implement longevity in their breeding programs they have to make several decisions. This thesis aims to give tools to make those decisions.
Chapter 2 gives an overview of the literature containing estimates of heritabilities of longevity traits and correlations between longevity and conformation traits. The results of Chapters 3 and 4 of this thesis are included as well. There are many different definitions of longevity. In this thesis, two distinctions are made: 1. between lifetime and stayability traits, and 2. between uncorrected and functional longevity traits. Lifetime traits measure the period a cow is alive or producing, and are usually expressed in days. Stayability traits measure whether or not a cow is alive at a certain point in time. Functional longevity traits are corrected for milk production, thus aiming to be a better measure for involuntary culling. In Chapters 1 and 7 of this thesis, residual longevity is introduced, which is longevity corrected not only for milk production but also for all other traits that are already in the breeding goal. So far, this trait has not been used in practice. From the literature it is concluded that, in general, heritability of longevity traits is below 0.10. The heritability of stayability traits is lower (around 0.04) than that of lifetime traits (around 0.09), and the heritability of functional longevity traits is lower (around 0.07 for lifetime traits and around 0.03 for stayability traits) than that of uncorrected longevity traits. Genetic correlations among different longevity traits are generally strong. Genetic correlations between longevity and conformation traits are strongest for conformation traits describing the mammary system and, to a lesser extent, feet and legs. The reliability of a breeding value prediction of a sire based solely on the conformation information of his daughters is approximately 55% at maximum.
In Chapter 3, the longevity realized of cows born in different years (1978 through 1985) has been calculated. Longevity of cows born in 1978 through 1984 decreases, and longevity of cows born in 1985 is at the same level as the longevity of cows born in 1978. In 1984, the quota system was implemented in the Netherlands and farmers culled 20% more cows than their normal annual culling percentage. These cows, of course, were born before 1984. Besides this process, during the eighties large-scale crossing with Holstein-Friesian bulls took place. The original Dutch-Friesian cow population was replaced by Holstein-Friesians, and this process was accelerated by imlementation of the quota system. Both processes not only affected longevity of dairy cows realized in the Netherlands, but also the estimates of heritabilities. Data on cows born in 1978, 1982, or 1985 were used to estimate heritabilities, and the estimates were highest for the 1978 dataset, lower for the 1982 dataset, and lowest for the 1985 dataset. Possible explanations are that the population was under strong selection during the period considered, that the genetic background of the population changed, and that under the quota system, farmers base their culling decisions on a shorter planning horizon, thus increasing the environmental variation of longevity traits.
In Chapter 4, data on cows born in different years (1978, 1982, and 1989/1990) were used to estimate genetic correlations between longevity and conformation traits. These parameters were also affected by the changing population structure during the eighties. In the 1978 data file, the correlation between functional herdlife and type was rather weak (0.16) while in the 1982 data file, this correlation was very strong (0.46). For the 1989/1990 data file, only stayability traits could be analysed because cows had not had enough time to be culled. The correlation between functional stayability until 48 months of age and type was 0.21. The strongest correlation was between functional stayability and the subjective score for udder (0.93), followed by the subjective score for feet and legs (0.43). The estimate of 0.93 is probably too high but also from other studies it was concluded that, apart from production, the udder is the most important factor determining longevity of a dairy cow. From Chapters 3 and 4 it was concluded that especially in an upgrading population estimates of genetic parameters should be based on the most recent data possible, and that estimation of these parameters should be repeated regularly.
In Chapter 5 the value of a relatively new method in animal breeding was investigated: survival analysis. Survival analysis differs in two aspects from traditional methods of analysis: 1. it correctly utilizes information from censored records, i.e., records of cows that are still alive at the moment of data collection; and 2. effects can be modelled in a time-dependent way, yielding a more realistic model. Breeding values of sires for longevity were estimated in three different ways: as the average realized longevity of the sire's daughters, with a best linear unbiased prediction, and with survival analysis. This was done using data from small and from large farms to identify a possible genotype by environment interaction. The phenotypic average of the sire's daughters had weak rank correlations with the other two methods of breeding value prediction (ranging from -0.32 to 0.46). The correlation between the best linear unbiased prediction and the survival analysis prediction was strong (-0.91 and -0.94 on small and large farms, respectively) if only uncensored records were used in the survival analysis, and weaker (-0.71 on both small and large farms) if censored records were included as well. Correlations were negative due to the definition of the traits: in the best linear unbiased prediction the length of productive life was analysed, and in the survival analysis the risk of being culled. A long length of productive life is associated with a small risk of being culled. Thus it was concluded that best linear unbiased prediction and survival analysis mainly differ by the data that can be included in the analysis. No different rankings of sires on small or large farms were found with any of the three methods. From the survival analysis, it appeared that cows with a high percentage of Holstein-Friesian genes had a lower chance of being culled than cows with a low percentage, confirming the hypothesis in Chapters 3 and 4.
Even though censored records can be analysed as well in survival analysis, a certain number of uncensored data is needed for a reliable breeding value prediction. Young bulls will probably not have a sufficient large number of daughters that have already been culled. Thus, conformation traits might be used for an early breeding value prediction, because they have reasonably strong correlations with longevity and can be measured early in a cow's life. In practice, a breeding value prediction will contain parental information on longevity, direct information on longevity of a sire's daughters, and indirect information on conformation of a sire's daughters. In Chapter 6 survival analysis was used to investigate the importance of conformation traits for the risk of a cow to be culled. This risk was corrected for milk production. Both the phenotypes of the cows themselves and their sires' breeding values for conformation were included in a model. The cows' phenotypes explained more variation in the risk of being culled than their sires' breeding values. In general, smaller cows with a steep rump angle, shallow udder, high score for udder and for feet and legs had the lowest chance of being culled. Survival analysis was also used to predict breeding values of sires for longevity based solely on the longevity of their daughters. These breeding values were correlated with the sires' national proofs for conformation traits, to obtain approximations of genetic correlations. The correlations were strong for nearly all conformation traits except height, rear legs set, and size. In the national proofs the conformation traits were not corrected for each other, while in the survival analysis they were.
In Chapter 7 it was argued that survival analysis should be used whenever possible to predict breeding values for longevity, even though with current computer capacities only a sire model can be used. Choosing this method implies that a lifetime trait has to be analysed. If length of productive life is analysed, a Weibull model can be assumed, which simplifies the calculations. In practice, this breeding value prediction will have to be combined with information on conformation to obtain a reliable breeding value for longevity early in a bull's life. Because most breeding programs of dairy cows pay already much attention to milk production, functional longevity will be more informative for breeding decisions than uncorrected longevity.