Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Economic efficiency of Canadian and New Zealand sires in Canadian and New Zealan dairy herds and its… Charagu , Patrick Kang’ethe 1997

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
831-ubc_1998-27117X.pdf [ 5.28MB ]
Metadata
JSON: 831-1.0088670.json
JSON-LD: 831-1.0088670-ld.json
RDF/XML (Pretty): 831-1.0088670-rdf.xml
RDF/JSON: 831-1.0088670-rdf.json
Turtle: 831-1.0088670-turtle.txt
N-Triples: 831-1.0088670-rdf-ntriples.txt
Original Record: 831-1.0088670-source.json
Full Text
831-1.0088670-fulltext.txt
Citation
831-1.0088670.ris

Full Text

ECONOMIC EFFICIENCY OF CANADIAN AND NEW ZEALAND SIRES IN CANADIAN AND NEW ZEALAND DAIRY HERDS AND ITS RELATIONSHIP WITH OTHER TRAITS by PATRICK KANG'ETHE C H A R A G U B.Sc.(Agric), The University of Nairobi, 1987 M.Sc. (Animal Breeding and Genetics), The University of Nairobi, 1992  A THESIS SUBMITTED IN P A R T I A L F U L F I L L M E N T OF T H E REQUIREMENTS FOR T H E D E G R E E OF D O C T O R OF PHILOSOPHY  in  T H E F A C U L T Y OF G R A D U A T E STUDIES (Department of Animal Science)  We accept this thesis as conforming ___tothe requireTTstendard  T H E U N I V E R S I T Y OF BRITISH C O L U M B I A November, 1997 © Patrick Kang'ethe Charagu  In  presenting this  degree  at the  thesis  in  University of  partial  fulfilment  of  of  department  this thesis for or  by  his  or  scholarly purposes may be her  representatives.  permission.  Department The University of British Columbia Vancouver, Canada  for  an advanced  Library shall make it  agree that permission for extensive  It  publication of this thesis for financial gain shall not  DE-6 (2/88)  requirements  British Columbia, I agree that the  freely available for reference and study. I further copying  the  is  granted  by the  understood  that  head of copying  my or  be allowed without my written  ABSTRACT  The main objectives of this study were to test for differences in the economic efficiency of daughters of Canadian and New Zealand dairy sires born in herds in either country and to determine the relationships between economic efficiency and first lactation production traits. The data used were extracted from lifetime records of a planned trial involving the mating of 20 proven sires selected from New Zealand and 20 proven sires selected from Canada to cows in 20 New Zealand and 10 Canadian herds. The data consisted of 834 lactation records from 343 cows in Canada and 3621 lactation records from 834 cows in New Zealand. The first stage in the study involved choosing the best growth function for describing the growth curve of each cow and this was used to estimate the liveweight at each calving. The liveweights were then used to estimate feed requirements, which were in turn used in the computation of the economic efficiency measure, Returns/Feed Requirements, for different production periods. The pricing systems used in the provinces of British Columbia and Ontario were considered when computing economic efficiency in Canada. Heritability estimates of the economic efficiency traits, growth and production traits were estimated using animal model DFREML as were the genetic and phenotypic correlations between these in New Zealand.  The existence of genotype by  environment interaction for the various traits was also tested. Heritability estimates for economic efficiency traits across both countries ranged from 0.12  to 0.34.  The genetic correlations between first lactation yield traits and economic  efficiency traits in New Zealand ranged from 0.44 to 0.70.  Genetic correlation between  economic efficiency in the first lactation and lifetime economic efficiency was 0.73.  In New Zealand daughters of New Zealand sires had significantly higher lifetime economic efficiency than daughters of Canadian sires. Daughters of both strains did not differ significantly for all other economic efficiency traits in either environments. A significant genotype by environment interaction was found at the macro level for first lactation protein yield and percentage protein. Among economic efficiency traits a significant interaction was only realized for lifetime economic efficiency with British Columbia prices. There was no interaction for any of the other economic efficiency traits and none for mature liveweight. Significant genotype by environment interactions were realized at the micro level for all economic efficiency traits for both strains of sire and for first lactation yield traits among Canadian sires.  iii  T A B L E OF CONTENTS  ABSTRACT  ii  T A B L E OF CONTENTS  .  LIST OF T A B L E S  iv vii  LIST OF FIGURES  ix  ACKNOWLEDGEMENTS  x  DEDICATION  xi  C H A P T E R 1: G E N E R A L INTRODUCTION  1  1.1 Dairy production in New Zealand  1  1.2 Dairy production in Canada  3  1.3 The Canadian and New Zealand Holstein-Friesian strains  4  1.4 Objectives  5  1.5 R E F E R E N C E S  9  C H A P T E R 2: G R O W T H C U R V E A N A L Y S I S FOR D A U G H T E R S OF C A N A D I A N A N D N E W Z E A L A N D SIRES IN C A N A D I A N A N D N E W Z E A L A N D D A I R Y HERDS 10 2.1 S U M M A R Y  10  2.2 INTRODUCTION  11  2.3 R E V I E W OF R E S E A R C H O N G R O W T H FUNCTIONS  13  2.3.1 Genetic parameters for growth curve parameters 2.4 M A T E R I A L S A N D M E T H O D S  18 21  2.4.1 Data  21  2.4.2 Fitting of growth curve functions  22  2.4.3 Comparison of growth function traits  24  2.4.4 Comparison of actual liveweights  25  2.5 R E S U L T S A N D DISCUSSION  26  2.5.1 Goodness of fit and comparisons among the curves  iv  26  2.5.2 Heritability estimates for curve parameters  30  2.5.3 Factors affecting growth curve parameters  31  2.5.4 Comparison of observed live weights  36  2.6 CONCLUSIONS  41  2.7 REFERENCES  44  CHAPTER 3: GENETIC ANALYSIS OF ECONOMIC EFFICIENCY OF DAUGHTERS OF CANADIAN AND NEW ZEALAND SIRES IN CANADIAN AND NEW ZEALAND DAIRY HERDS 46 3.1 SUMMARY  46  3.2 INTRODUCTION  47  3.3 REVIEW OF RESEARCH ON PROFIT FUNCTIONS IN DAIRY CATTLE  49  3.3.1 Different measures of profitability  51  3.3.2 Identification of economically important traits  56  3.3.3 Other factors affecting profitability  62  3.3.4 Genetic parameters for profitability  63  3.4 MATERIALS AND METHODS  64  3.4.1 Data  64  3.4.2 Definition of the economic efficiency trait  65  3.4.2.1 Estimating food (Metabolizable Energy) requirements  67  3.4.2.2 Liveweights  70  3.4.2.3 Returns  71  3.4.3 Analyses  •  73  3.5 RESULTS AND DISCUSSION  .75  3.5.1 First lactation yield traits  75  3.5.2 Economic efficiency and lifetime traits  81  3.5.3 Heritability estimates 3.5.4 Genetic and phenotypic correlations (New Zealand)  ,  84 88  3.6 C O N C L U S I O N S  91  3.7 R E F E R E N C E S  94  C H A P T E R 4: G E N O T Y P E B Y E N V I R O N M E N T INTERACTION FOR E C O N O M I C E F F I C I E N C Y A N D CONSTITUENT TRAITS  99  4.1 S U M M A R Y  99  4.2 INTRODUCTION  100  4.3 R E V I E W OF R E S E A R C H O N G E N O T Y P E B Y E N V I R O N M E N T I N T E R A C T I O N . . 101 4.3.1 Studies on genotype by environment interaction in dairy cattle  104  4.4 M A T E R I A L S A N D M E T H O D S  107  4.5 R E S U L T S A N D DISCUSSION  110  4.5.1 Genotype by environment interaction at the macro level  110  4.5.2 Genotype by environment interaction at the micro level  116  4.6 CONCLUSIONS  118  4.7 R E F E R E N C E S  121  C H A P T E R 5: C O N C L U D I N G R E M A R K S  123  vi  LIST OF TABLES Table 2.1. The four growth functions  22  Table 2.2. Overall means and standard deviations of sums of squared errors (SSE) and of growth curve parameter estimates for the three growth functions in Canada. (n=255)  27  Table 2.3. Overall means and standard deviations of sums of squared error (SSE) and of growth curve parameter estimates for the three growth functions in New Zealand  27  Table 2.4. Heritability estimates of growth curve parameters in the New Zealand environment  31  Table 2.5. Analysis of variance (R x 100) results for the Von Bertalanffy function parameters in Canada  33  Table 2.6. Least squares means and standard errors for growth curve parameters estimated using the Von Bertalanffy function for cows in Canada  33  Table 2.7. Analysis of variance (R x 100) results for the Von Bertalanffy function parameters in New Zealand  34  Table 2.8. Least squares means and standard errors for growth curve parameters using the Von Bertalanffy function for cows in New Zealand  34  Table 2.9. Analysis of variance results (R x 100) for heifer liveweights at birth, 6, 12, 18 and 30 months of age in Canada  38  Table 2.10. Least squares means and standard errors of liveweights for heifers sired by Canadian and New Zealand bulls in Canada  38  Table 2.11. Analysis of variance results (R x 100) for heifer liveweights at birth, 6, 18 and 30 months of age in New Zealand  40  Table 2.12. Least squares means and standard errors of liveweights for heifers sired by Canadian and New Zealand bulls in New Zealand  40  Table 3.1. Analysis of variance results (R x 100) for first lactation and economic efficiency traits and calving interval in the Canadian environment  76  Table 3.2. Least squares means and standard errors by strain of sire for first lactation and lifetime traits in the Canadian environment  77  2  2  Table 3.3. Analysis of variance results (R x 100) for first lactation and economic efficiency traits and calving interval in the New Zealand environment 2  Table 3.4. Least squares means and standard errors for first lactation and lifetime traits in the New Zealand environment  vii  ..79  80  Heritability estimates (and standard errors) for first lactation and economic efficiency traits in Canada and New Zealand  85  Phenotypic variances for first lactation and economic efficiency traits in Canada and New Zealand  86  Estimates of genetic correlations (above diagonal) and phenotypic correlationsfbelow diagonal), between first lactation production traits and economic efficiency traits in the New Zealand environment (standard errors of genetic correlations given in brackets)  89  Analysis of variance (R x 100) results for first lactation yield traits, weight traits and economic efficiency traits from model 4.1  Ill  Least squares means for first lactation yield traits, mature weight and economic efficiency traits for daughters of Canadian and New Zealand sires in Canadian herds  112  Least squares means for first lactation yield traits, mature weight and economic efficiency traits for daughters of Canadian and New Zealand sires in New Zealand herds  113  Expected correlations [E(r)] and product moment correlations (R) of sire breeding values for weight, lactation one and economic efficiency traits  117  Table 3.5.  Table 3.6.  Table 3.7.  Table 4.1.  Table 4.2.  Table 4.3.  Table 4.4.  viii  LIST OF FIGURES  Figure 2.1. Mean Growth curves fitted by the Von Bertalanffy function for cows sired by the Canadian and New Zealand bulls in the Canadian and New Zealand environments Figure 4.1.  Graphical illustration of genotype by environment interaction  ix  37 102  ACKNOWLEDGMENTS  I wish to express my great gratitude to my supervisor Dr. Ray Peterson for his advice, creative criticism and guidance throughout the period of my study.  I also wish to thank the other members of my supervisory committee, Drs., G. Namkoong, R. M . Tait, K . M . Cheng and J. Vercammen for their comments and suggestions on this thesis.  M y gratitude to the Canadian International Development Agency (CIDA) for awarding me the scholarship, through the University of British Columbia/University of Nairobi Linkage Project, that made this study possible.  M y heartfelt appreciation and thanks to my wife, Wakabari, and daughter, Wanjiku, for their encouragement, love and understanding in the course of my program. Their belief in me was a source of challenge, and at the same time courage and strength.  I cannot forget my mother, Hannah Wanjiku wa Charagu, my first ever teacher, at home and at school. Y o u sowed the seed. To my father, Charagu "Witu", for his great pride in us, his children.  This has always made us strive even harder.  I also wish to acknowledge the  encouragement and support of my siblings, Kimani (Richman), Wangui, Wanja, and the late Wanjiku.  Special thanks too to my parents, brothers and sisters in-law for their continued  encouragement and prayers.  Dedicated to my late sister, Jane Wanjiku, and my children.  CHAPTER 1: GENERAL INTRODUCTION  1.1. Dairy production in New Zealand The New Zealand dairy industry is small by world standards, with about 2.2 million cows producing about 7.4 million tonnes of milk, which is about 1.5% of the world's production (Bryant 1993). The industry has its roots in the Shorthorn breed which was replaced by the Jersey with the introduction of artificial insemination in the 1950s. The Friesian in turn replaced the Jersey as the dominant breed 20-30 years later. Most herds are between 100 and 200 cows, with 80% of herds being less than 200 cows (Bryant 1993). The average size of a dairy farm is 70 hectares (Guy 1993). Some farms are dedicated to only producing milk for local consumption while others produce for the export market.  The latter are referred to as factory supply or  seasonal farms and account for approximately 90% of the milk produced in New Zealand. B y world standards, the factory supply herds have, on average, a large herd and low production per cow (Bryant 1993). They have high milkfat test results due to the type of cows used and the management under which they are kept. The two main breeds are the Friesian and Jersey, with crossbreds between the two being very common. The prices farmers receive for their milk are almost entirely dependent on export prices which are lower than what can be obtained on internal markets and have indeed consistently declined over the years. To remain profitable, farmers have had to devise production systems that are lower in costs than those of dairy farmers in other countries. This has been achieved in two ways; 1)  A minimal use of expensive inputs like high energy supplements, labour,  machinery, and use of nitrogenous fertilizers, 2) almost complete dependence on pasture, the species of which are mostly perennial ryegrass and white clover (Guy 1993). The end result is that New Zealand has the lowest cost production of the major dairy industries world wide  1  (Murphy 1993). For many years in the past farmers were paid for kilograms of fat delivered. Due to the increasing importance of milk protein products in many countries, a new payment system incorporating the amount of protein and the amount of milk fat has been introduced in the past few years. The dependence on pasture alone makes feed supply highly seasonal. About 60% of the annual pasture is available during the 5 months of September to January when the growth rate increases 4 to 5 times the rate in July (Bryant 1993).  The dairy farmers overcome this  seasonality of pasture by making milk production seasonal. Calving takes place at the start of spring (around September and October) and drying off in the autumn. To take advantage of the seasonality of pasture a high level of reproductive efficiency is required. On average every cow in the herd must calve every 365 days and it is imperative that the whole herd calves in a period of 6-8 weeks. Failure to get in calf is a major reason for culling (Murphy 1993) in the seasonal herds.  Calving late in the "calving window" reduces overall performance because it reduces  days in milk and fails to make full utilization of the spring flush of pasture growth. Calving too early may result in a longer lactation, but also could mean severe underfeeding in early lactation leading to low total lactation yield. The average production per cow in New Zealand is approximately 3200 litres with 150 kg milk fat (4.72%) and 115 kg protein (3.63%) (Guy 1993) per lactation. Murphy (1993) argued that New Zealand cows are underachievers in terms of production per cow. He reasoned that the main cause of that is the underfeeding of cows at particular times of the year, and the other being extremely low heifer liveweights at calving.  The consequence is that in New  Zealand the genetic potential of the cows is not fully exploited. The New Zealand dairying system is a severe production environment the outcome of which is the relatively low production per cow.  2  1.2. Dairy production in Canada  The Canadian dairy industry, as opposed to that of New Zealand, is a high input and high performance system. All the herds are kept under intensive management systems and although some herds practice some grazing during the summer months after the pastures have been harvested, total confinement is more predominant. The average Canadian herd includes 60 cows and heifers over 2 years of age (Christensen and Fehr, 1993). Most of the smaller herds are housed in tie stall (stanchion) bams while most larger than average herds are kept in freestall bams. The cows are reared and kept under high energy and high nutrient-density diets with heavy reliance on conserved forages and cereal grains. Forages include com silage in parts of Ontario, Quebec and British Columbia with a greater reliance on cereal silage, mainly from barley, in much of western Canada (Christensen and Fehr, 1993). Alfalfa hay and silage and a number of types of grass hay are also utilized throughout Canada. The feedstuffs used in the various regions will vary due to the differences of crops and pasture species that thrive and are cheaper in the regions. Calving and production take place the year round and consequently the heavy dependence on conserved forages. Holstein, the predominant breed, produces an average of 7,717 kg of milk per lactation while the weighted average of all breeds is more than 7,500 kg in a 305 day lactation (Christensen and Fehr, 1993). The Canadian dairying industry is labour and capital intensive and this makes it one of the highest cost dairy industries in the world. Murphy (1993) compared the cost of milk production and found that it was 3 times more in Canada and the US than in New Zealand. Only Denmark was higher with a cost 3.33 times more than that in New Zealand. The industry in Canada is highly mechanized in addition to being labour and capital intensive, all of which result in high production costs. Due to this high cost of producing milk, the prices received by farmers are also high compared to those paid to the dairy farmers in New Zealand. This is because in the  3  computation of prices, the costs of production are also considered as is the quota assigned to each farmer.  1.3. The Canadian and New Zealand Holstein-Friesian strains The dominant breed among dairy cattle in Canada is the Holstein making up approximately ninety percent of the dairy population (Christensen and Fehr, 1993). Over the years, each country has pursued its own breeding and selection policies and goals with little exchange of germplasm, particularly New Zealand to Canada. In New Zealand the Friesians were increasingly selected for milkfat yield and in recent years with the advent of compensation for protein content, for milkfat and protein yields. In Canada on the other hand, until payment based on milk components started, most of the selection over many years was done for increased milk production with weight given to milkfat differential. This was possibly so since most of the milk was targeted for the internal, rather than the export market, as opposed to New Zealand. On average the Canadian Holstein-Friesian is also bigger and heavier than the New Zealand strain (Bar-Ananef al, 1987). In addition to being selected for higher milkfat (and protein) yields, the New Zealand strain of Holstein-Friesian has also been selected towards these goals under a "harsher" system of production. The New Zealand system is considered a "harsher" environment since dairying is done under a lower plane of nutrition with little supplementation from conserved forages or high nutrient-density concentrates. The Holstein-Friesian in New Zealand has therefore had to adapt to a relatively different production environment, from that under which selection in Canada has been carried out. Another aspect of the selection of the Holstein-Friesian in New Zealand is that compared to other countries there has been a high selection intensity on the bull-to-cow and the  4  dam-to-bull path (Cunningham, 1983). It is the goals of selection, however, rather than the means of achieving them that leads to differences in strains. The goals in both Canada and New Zealand were discussed earlier and based on these goals and the production environment under which selection has been done, one can argue that selection may have resulted in two different strains. Researchers have in the past investigated the existence of differences between various strains of the Holstein-Friesian (Jasiorowski et al. 1983; Bar-Anan et al. 1987; Graham et al., 1991; Peterson, 1991) and come to the conclusion that differences do exist between some of the strains.  1.4. Objectives The current study will try to test for differences in the economic efficiency of the Canadian and New Zealand strains of the Holstein-Friesian breed in both Canada and New Zealand in order to determine which of the two strains, i f any, is more efficient since this will then give an indication as to which is more profitable. Economic efficiency (Returns/Food Requirements) in this case is used as an indicator of profitability since there wasn't enough information to compute profit itself in the standard form of Returns - Costs. The two (economic efficiency and profit), it will be shown later in section 3.3.2, are linearly related and thus efficiency is used as a form of index for profitability. The idea is to first compute the costs of feeding in terms of energetic requirements.  To achieve this the weights of the cows are  necessary, and the first part of the study will chart the growth curves of the cow so that the maintenance energy requirements can be computed for various points in the course of their herdlife. The second part of the study delves into the computation and comparison of the  5  economic efficiency of the cows which also takes into account the production energy requirements which are computedfromthe production records. In so doing both the strains and environments are assessed with reference to each other in terms of cow and sire economic efficiency. The existence of genotype by environment interaction is also addressed. The issue of genotype by environment interaction is important because it enables both breeders and commercial producers to know whether the breeding values (for both production and economic efficiency) of sires from progeny testing of their daughters in one of the environments can be utilized in the other environment. The lack of a genotype by environment interaction means that breeding values in one environment can be adapted in the other while the existence of the same means breeding values are not transferable and sires would have to be tested in both environments. Thus the overall objectives are; 1) To determine the best growth function for describing the growth curves of the study cows, to determine the factors affecting these growth curves and to compare the two strains for their growth curves and observed liveweights at various ages. 2) To determine and compare lifetime and partial lifetime economic efficiency of daughters of Canadian and New Zealand sires in both countries. 3) To determine the relationship between measures of lifetime and partial lifetime economic efficiency of dairy cows, and the relationship between the two and other cow traits. The traits to be considered are those measured early in life and thus determine the potential of using such traits in predicting sire economic efficiency. 4) To investigate the existence of a genotype (strain) by environment (GxE) interaction in yield, weight and economic efficiency traits in the two countries.  6  Lifetime records of the daughters of selected New Zealand and Canadian sires recorded in a planned mating program (the Canadian/New Zealand genotype by environment interaction trial (CANZ)) will be used in this study. The trial was designed to test for the existence of genotype by environment interaction for milk yield and milk component traits in the Canadian and New Zealand production environments. This was a planned trial which involved mating 20 proven sires selected from New Zealand and 20 proven sires selected from Canada to cows in 20 New Zealand and 10 Canadian herds. The constraints to selection of project sires was that they were widely used in the country of origin and had accurate official progeny tests. In New Zealand the mating plan was such that 10 bulls from each country were used in each herd in a structured design to ensure sires and herds were not confounded. In Canada, all the bulls were given the opportunity to sire daughters in all the project herds. The project sires were selected based on 1984 official progeny test information. This was accomplished by selecting available sires with the highest production proofs and conformation proofs that were average or above for mammary system and final classification.  The selected Canadian sire group was approximately 1.1 and 0.7 standard  deviations above the mean proof of the 254 "active" A I proven sires (spring, 1985) for milk and fat respectively. The selected New Zealand sire group was 0.8 and 1.0 standard deviations above the mean proof of the 184 available "active" sires (1983-1984 sire summary) for milk and fat respectively. In New Zealand nineteen commercial herds and the number four herd at Massey University were the original cooperators in the planned mating program for a total of 20 herds. These herds were recruited by the New Zealand Dairy Board (Livestock Division) from the factory supply herds which had a long history of milk recording and a high quality of record keeping.  A l l herds in New Zealand relied on pasture grazing for most of their feeding  7  management with little and mostly no supplementation (apart from mineral licks and other vitamin supplements). The herds relied principally on intensive pasture management programs rather than conserved forages and concentrate feeds as was the practice among the Canadian herds. Seasonal calving was practiced and this was determined by herd experiences on the seasonal availability of pasture. Cows were dried depending on the availability of pasture with groups of cows being dried at one time without consideration of calving date or days milked. The 20 New Zealand herds involved in the trial were from the 5 major dairy regions on the North Island. The 10 Canadian herds involved in the project were the Agriculture Canada herd at Agassiz in British Columbia, University of Guelph, University of Manitoba, University of British Columbia (Oyster River and South Campus herds), McDonald College, Nova Scotia Agricultural College and Olds Agricultural College. In Canada, the cows were mainly fed on conserved forages and supplemental concentrates with limited grazing practiced in some of the herds.  8  1.5. REFERENCES  Bar-Anan, R., Herman, M., Ron, M. and Weller, J. I., 1987. Comparison of proven sires from five Holstein-Friesian strains in high-yield Israeli dairy herds. Livest. Prod. Sci. 17: 305322. Bryant, A. M., 1993. Dairying in New Zealand. An international perspective of dairying in New Zealand. Proceedings of the XVII International Grassland Congress 1993: 1587-1588. Christensen, D. A. and Fehr, M. I., 1993. The Dairy Industry. In, Animal Production in Canada: pp 75. Publ. University of Alberta, Faculty of Extension. Ed. J. Martin, R. J. Hudson and B. A. Young. Cunningham, E. P., 1983. Structure of dairy cattle breeding in Western Europe and comparisons with North America. J. Dairy Sci., 66: 1579-1587. Graham, N. J., Burnside, E. B., Gibson, J. P., Rapitta, A. E. and McBride, B. W., 1991. Comparison of daughters of Canadian and New Zealand Holstein sires for first lactation efficiency of production in relation to body size and condition. Can. J. Anim. Sci. 71: 293300. Guy, B., 1993. Dairy farming in New Zealand. Proceedings of the XVII International Grassland Congress 1993: 995-997. Jasiorowski, H., Reklewski, Z. and Stolzman, M., 1983. Testing of different strains of Friesian in Poland. I. Milk performance of Fj paternal Friesian strain crosses under intensive feeding conditions. Livest. Prod. Sci. 10: 109-122. Murphy M., 1993. An international perspective of dairying in New Zealand. Proceedings of the XVII International Grassland Congress 1993: 1585-1586. Peterson R., 1991. Evidence of a genotype/environment interaction between Canadian Holstein and New Zealand Friesian cattle under Canadian and New Zealand management systems. Proc. of the 42nd Annual Meeting of the European Association for Animal Production, Berlin, Sept. 9-12, Vol. 1: pp 49.  9  CHAPTER 2: GROWTH CURVE ANALYSIS FOR DAUGHTERS OF CANADIAN AND NEW ZEALAND SIRES IN CANADIAN AND NEW ZEALAND DAIRY HERDS  2.1. SUMMARY  Four growth functions, the Brody, the Gompertz, the Logistic and the Von Bertalanffy, were compared for their ability to describe the growth curves of cows born in the Canadian New Zealand Trial in both Canada and New Zealand. Comparisons of means of sums of squared error showed that on average the Von Bertalanffy had the lowest mean and was therefore judged to have the best goodness of fit to the weight data. It would therefore be used to estimate feed requirements in the next chapter. The heritabilities for the three curve growth parameters; A, which represents asymptotic mature weight, b - a measure of the proportion of mature weight at birth and k, - a maturation index, were estimated for all four functions for the cows bom in New Zealand. Heritability estimates for mature weight using the four functions were similar at approximately 0.43, while those for b ranged from 0.26 to 0.35 and for k 0.16 to 0.26. The daughters of the two strains of sires, Canadian and New Zealand were compared for their parameters of the Von Bertalanffy function. In both environments Canadian sired cows were significantly heavier at maturity (A). There was no significant difference between the two strains, however, for the b and k parameters in both environments. Sire within strain of sire was important for all three parameters in New Zealand but not for any of them in Canada. The strains were also compared for their observed liveweights at birth, 6, 12 (Canada only), 18 and 30 months. In New Zealand daughters of Canadian sires were significantly heavier at all ages while in Canada no significant difference was observed at any of the age points.  10  2.2. INTRODUCTION  In research that is geared towards assessment of economic efficiency or profitability in animals one of the most important variables considered is feed intake. In the absence of feed intake information, however, it is possible to estimate feed requirements based on liveweight and production level, where production could be in the form of milk, meat, eggs or any other form of product. Given the importance of liveweights in such studies one of the aspects is that the weights of the animals under trial are taken regularly over the experimental or desired period. In those cases where weights needed at particular instances are not available, extrapolations and interpolations could be relevant solutions. Once the weight and production profiles are known, feed requirements can be estimated and it then becomes possible to compute estimates of economic efficiency. In this study feed intake information was not available and hence there was a need for a procedure to estimate weights at required points. Extrapolations and interpolations are only meaningful when the nature of the weight changes with time is understood. One of the ways this could be achieved is through the use of growth curves. Growth curves are a reflection of the lifetime interrelationships between an individual's inherent impulse to grow and mature in all body parts and the environment in which these impulses are expressed (Fitzhugh, 1976). Understanding of growth curves is therefore a crucial facet in the research in, and recommendations made on, lifetime production and efficiency.  One way of understanding and utilizing growth curves is by quantifying or  describing them through the use of mathematical functions which are often referred to as growth functions. Growth functions apart from describing growth curves are used for many other purposes, among which are; 1) to summarize descriptively how a measurement changes with time in an  11  animal or a group of animals, 2) to provide a criterion for comparing animals in order to pass judgment on their performance, 3) to compare the consequences of different feeding or other management treatments in experiments, to estimate for individual animals either the value that a variate had at an age when no measurement was made or the age at which a variate achieved a specified value, 4) to predict the value of a variate at some future date or to predict the ages at which individual animals will achieve specified measurements (Finney, 1978). The objectives of the current study were; (i) To determine the best growth function, that is the one with the best fit for the weight data of daughters of Canadian and New Zealand bulls in Canadian and New Zealand herds. This function will then be used to provide individual animal weights for the estimation of feed requirements, (ii) To describe the growth curves of individual cows in the CANZ trial and to determine the factors affecting these growth curves, (iii) To compare the two strains for their growth curves and observed liveweights at various ages. The null hypotheses to be tested are that the different growth functions do not significantly differ in their goodness of fit for the actual growth curves, and that the cows sired by bulls from the two different strains will show no differences in their growth traits, curves and hence their parameters.  12  2.3. REVIEW OF RESEARCH ON GROWTH FUNCTIONS  Liveweight is a variable that can be measured repeatedly during the lifetime of an animal, that is, it is a time series variable. From birth to maturity or death, weight follows a continuous curve, although, practically, only a few points over an animal's lifetime are usually measured. When approximated, the growth curve is smooth, steadily increasing to maturity, and subsequently nearly constant or perhaps increasing or decreasing slightly. Under closer examination however, irregularities are observed. These are in the form of small waves within each short time period imposed upon the general trend, growth spurts, and negative effects of short term illnesses and adverse environmental or physiological occurrences. Fluctuations are also observed in female animals due to the effects of pregnancy and lactation. If such curves are averaged over animals within a population (for example over one breed on a given feeding regime), the irregularities disappear and a simpler "typical" curve appears. It is essential though to distinguish between individual growth curves and the idealized mean curve for a population or a treatment group (Finney, 1978). A growth curve representing liveweight, (and any other variate closely related to the general size of an animal) usually has three major phases, an initial slow but steady acceleration, a middle period of rapid growth that may look approximately linear, and a closing phase in which growth slows towards a mature (asymptotic) level. Nonlinear mathematical functions can be fitted to such growth curves to describe them and in so doing a set of parameters related to growth can be derived. These parameters can be interpreted biologically and also used to derive other growth traits of interest. The genetic relationships of these parameters may also be of interest. Selection for other unmeasured growth characteristics would then become more feasible once they have been estimated by use of the growth functions.  13  The most familiar growth functions, as was later shown by Richards (1959), are essentially specialized cases of one general four parameter function describing the pattern of growth that has come to be known as the Richards function. The Richards function is described as; W =A(\-be- ') k  M  t  where W = body weight at age t (days) t  A = asymptotic (mature) body weight (kg) b = constant of integration k = maturing index M = weight exponent constant  Although A is denoted as the mature weight, this does not imply that A is the heaviest weight obtained by the individual, but it indicates the average weight of the cow at maturity, independent of short-term fluctuations in weight. The parameter b indicates the proportion of the asymptotic mature weight to be gained after birth, established by the initial values of W and t. This parameter adjusts for the situation in which W (initial weight) and/or to (time of origin) do 0  not equal 0. When W = t = 0, then b = 1. A small value of b represents a higher proportion of mature weight at birth. The parameter k is a constant that expresses the rate at which a logarithmic function of W, specific for each of the non-linear equations, changes linearly with time. It can be interpreted biologically as a maturation index, establishing the earliness with which the weight at any phase of growth, W, approaches the asymptotic mature weight A. The parameter M determines the proportion of the asymptotic mature weight at which the growth inflection point occurs. It is therefore responsible for the shape of the curve generated by the  14  function and can be called the form parameter. Weight at point of inflection (Wj) is estimated as;  so that; M = 1 gives WT = OA M = - l gives Wj. = .50A M= 3 gives W! = .30A M —» oo gives W[ - .37A  The nonlinear functions used to describe growth and which are, in essence, different expressions of the Richards formula include the Brody, also sometimes referred to as the monomolecular function, (Brody, 1945), the logistic (Pearl and Reed, 1923), the Von Bertalanffy (Von Bertalanffy, 1957) and the Gompertz (Winsor, 1932). M is fixed at 1 for the Brody, -1 for the logistic, 3 for the Von Bertalanffy and approaches infinity for the Gompertz. The four functions then have the following forms;  Gompertz:  -be  W, = Ae  Brody (Monomolecular): Von Bertalanffy: Logistic:  15  All the above four are three parameter functions. In theory the four parameter Richards function is superior to the others for fitting growth curves. Practically however, fitting of this function is made difficult by the number of parameters and lack of convergence in the M parameter (Brown et al, 1976; Bakker and Koops, 1978) Lack of convergence is especially seen in situations where observed weight records are scarce. It has therefore been easier to obtain results using the other forms of the function where the values of M, though different have already been assigned. The various growth functions have been used by researchers to fit the growth curves of different species of animals, ranging from mice to cattle. Eisen et al. (1969) fitted different growth functions to the growth data of mice. Barbato (1991) used the Gompertz function to describe the growth curves of poultry and Doren et al. (1989) used the Brody function to estimate parameters of the growth curves of beef bulls. The different functions might show different degrees of goodness of fit in different species or breeds of animals and in different environmental circumstances. Goodness of fit refers to the minimization of the deviations of the observed data points from the corresponding points on the fitted curve. As mentioned above with the Richards function computational difficulties do also vary with the functions. This is determined by the nature of a specific data. The models tend to be sensitive to the sparseness of data. Depending on the data and the function fitted to it parameter estimates may be obtained that are not within the limits of biological feasibility. Many of the functions have been compared for goodness of fit using different breeds and in different environments. Bakker and Koops (1978) working with Dutch Friesian, British Friesian and Holstein-Friesian cows tested functions with different values of M to obtain the curve with the best fit. They substituted 8 different values of M ranging from -1 to 1 and  16  compared the functions based on the smallest value of residual variance. They found that the function with an M value of 1/3 (Von Bertalanffy) gave the best fit to the data. They therefore used this function for further research in the adjustment of growth curves for the effects of lactation and gestation. Lopez de Torre et al. (1992) compared the fitness of the Richards, Brody and Von Bertalanffy functions to describe the growth curves of Retinta beef cows. They found that Richards function required long computing time, failed to converge for a large number of cows and had huge coefficients of variation (CV) for the estimates of parameters A b and k. Brody and Von Bertalanffy on the other hand were similar in the number of animals showing convergence and the CVs for parameter estimates. The various functions also differ in their ability to estimate the different parameters, particularly the mature weight A and weights at different ages along the curve. Perotto et al. (1992) compared the Brody, Richards, Gompertz and logistic functions using cows of the Holstein and Ayrshire breeds and their cross in Canada. They reported that Richards function seemed to give the smallest residuals and thus the best fit. The Brody had a tendency to overestimate and the Gompertz to underestimate. They however noted, as has been noted by others researchers (Brown et al., 1976; Lopez de Torre et al, 1992) that fitting the Richards function was computationally more demanding and suggested that perhaps another function would give a better fit to the growth curves. Koenen and Groen (1995) working with Dutch Black and White heifers also compared different functions for best fit to growth data. They compared the Gompertz, Von Bertalanffy, Michaelis-Menten, logistic, and Richards functions.  Their results showed that the Von  Bertalanffy function gave the bestfitto the data. Their recommendation of the Von Bertalanffy function is in line with those of Bakker and Koops (1978) and Lopez de Torre et al. (1992).  17  Research has also shown that some functions give a better fit for weight-age data over certain age periods than others. Perotto et al. (1992) reported that the Gompertz and logistic functions overestimated weights at birth with wide margins (compared to the mean of the observed weights) while the Brody and Richards functions overestimate it with smaller margins. In the same study the logistic function underestimated weights taken late in life whereas the Brody, Richards and Gompertz gave good predictions. Brown et al. (1976) reported that the Von Bertalanffy, Gompertz and logistic functions consistently overestimated weights at early ages, and that the logistic underestimated mature weight. The Richards model, however seemed to fit the data reasonably over all ages. Lopez de Torre et al. (1992) working with Retinta beef cows reported that at 60 to 97 months of age (maturity), Brody's function's A parameter overestimated the average by 14.6 kg while the Von Bertalanffy's exceeded by 10.1 kg. One can therefore only be conclusive with a given dataset after testing the various functions for the suitability of each over the range of the data.  2.3.1. Genetic parameters for growth curve parameters  Even though growth traits and other related traits have received a lot of attention from researchers in terms of genetic evaluation, the literature on genetic parameters of growth curve parameters is sparse. This lack of literature is seen in both dairy and beef cattle breeds and in the dairy breeds this could be explained by the fact that rarely do dairy animals get weighed once they enter their production phase. A few researchers, however have reported results of genetic parameter estimation on growth curves. The parameters A and b have demonstrated medium to high heritability estimates. The k parameter on the other hand has shown variability in its estimates of heritability, from low to moderate.  18  De Nise et al. (1985) estimated growth curve parameters for Richards and Brody's function using beef cows. Their heritability estimates using Brody's function were .44 ± .27, .39 ± .27 and .20 ± .26 for the A, b, and k parameters respectively. Using the same function Brown et al. (1972) obtained estimates of A, b and k of .34 ± .25, .62 ±.34 and .33 ±.25 for Hereford cattle. These estimates were similar to other estimates of heritability for directly measured mature weight (Brinks et al, 1964).  Heritability estimates of Richards growth function  parameters are few for cattle. De Nise et al. (1985) reported estimates of .44 ± .27, .24 ± .26, .32 ± .27 and .21 ± .26 for the A, b, k and M parameters respectively for beef cows. Timon and Eisen (1969) estimated heritabilities for the Richards function parameters for mice selected and unselected for post weaning gain. Their heritability estimates were .66 ± .15, .53 ± .15 and .30 ± . 14 for A, M and k parameters respectively. The Von Bertalanffy function, although having been reported to have a relatively good fit in cattle weight data, has had few estimates of genetic parameters. In the study of Koenen and Groen (1996) of Black and White dairy heifers they reported heritability estimates of growth parameters estimated using the Von Bertalanffy function of 0.21 ± 0.07, 0.31 ± .08 and .02 ± .05 for the A, b and k parameters respectively. Their heritability estimate for b is quite similar to those in the other studies with other functions but the A and k heritabilities tend to be slightly lower. Whether this is the general case with this particular function is difficult to determine. In an effort to genetically characterize both growth functions and their parameters researchers have also estimated genetic and phenotypic correlation between the various parameters (Brown et al, 1972a; De Nise et al., 1985; Koenen and Groen 1996), though not with consistent results. Using Brody's function De Nise et al. (1985) reported negative genetic correlations between A and b and between A and k, while the genetic correlation between b and k was positive. The negative correlation between A and b indicated that the larger the mature  19  size the greater the proportion mature weight in birth weight. The negative correlation between A and k on the other hand indicates that cows smaller at maturity were earlier maturing. A positive correlation between b and k implies that cows bom at lighter weights were associated with earlier maturity. The results on correlations in De Nise et a/.'s (1985) study were similar to those of Brown et al. (1972a) who also found the genetic correlations between A and b, A and k and b and k were negative, negative and positive, respectively. The relationships between the growth curve parameters are not, however, consistent across functions. In the same study of De Nise et al. (1985) above but using Richard's function the genetic relationship between A and b and the b and k parameters had signs (+/-) reversed from what they were when the parameters were estimated using Brody's function. This means that some of these parameters can not be interpreted biologically as similar traits in both functions, and specifically, the b parameter in Richards function cannot be said to be a measure of relative weight at birth. Groen and Koenen (1996) using the Von Bertalanffy function found a positive genetic correlation between A and b (.19 ± .20) and a negative genetic correlation between A and k (-.82 ± .21) among Dutch Black and White dairy heifers. The high negative correlation between A and k indicate that heavier heifers mature more slowly. They found no genetic correlation between the b and k parameters.  20  2.4. MATERIALS AND METHODS  2.4.1. Data  Weight records from the Canadian New Zealand genotype by environment interaction trial (CANZ) referred to in Chapter 1 were used.  In Canada the heifers were weighed  approximately once a month from birth until first calving and depending on the herd, they were weighed in the course of the first and other lactations after that. Some herds stopped weighing their animals in the course of the first lactation while others continued doing so for two, three or four lactations. In these analyses any animal without at least one weighing after the first calving was deleted and in this respect there was one herd in New Zealand that did not weigh their heifers at this stage and it was therefore left out of the analysis. To reduce bias from the effects of gestation in the last trimester any weight observations within the last 100 days prior to a calving were omitted from the analysis. In Canadian herds the minimum number of weighings per animal for the 260 animals remaining in the analysis was 11. The number of cows with complete weight records in New Zealand were 677. In New Zealand each herd weighed their project heifers at birth, and at approximately 6, 18 and 30 months of age. The weight data points in New Zealand were increased by assuming a linear growth pattern from one weighing age to the next. Using the linear interpolation weights were estimated for the heifers every 2 months from birth to 30 months. This way all the animals had 16 weight points. This was done because estimates of growth curves obtained by using the weights at the four actual weighing points (birth, 6, 12 and 30 months) gave estimates of mature weights that were unrealistically low. The inclusion of the interpolated weight points gave estimates of mature weight and weight at 30 months that were consistent with observed weights in New Zealand (Bevin Harris, personal communication).  21  2.4.2. Fitting of growth curve functions  Thefirststep was to compare four growth functions to obtain the one giving the best fit to the data. The functions compared were the Gompertz, Brody, Von Bertalanffy and Logistic growth functions which are given in Table 2.1 below. Table 2.1. The four growth functions evaluated in the study. -kl  Gompertz:  W =  Ae  Brody:  W, =  A(l-be  Von Bertalanffy:  w  =  Logistsic  w  = A(l + be  t  t  t  be  A(l- be -kty  where; W = body weight at age t in days, t  A = asymptotic (mature) body weight in kilograms, b = constant of integration, indicates the proportion of the asymptotic mature weight to be gained after birth, k = maturation index, establishes the earliness with which the weight at any phase of growth, W, approaches the asymptotic mature weight A.  Each of these functions was fit to the serial weight data of each cow using the nonlinear (TvTLIN) procedure of SAS (SAS Institute Inc., 1990). In the procedure no first order derivatives of the parameters were specified and therefore the DUD or Secant method (Ralston and Jennrich, 1979) for estimating nonlinear parameters was used with the maximum number of iterates specified as 50. Convergence of the iterative procedure was reached whenever;  22  (SSEj - SSEj.!) / (SSEj + IO" ) < C, 6  where SSE is the residual sum of squares after fitting the relevant function to the data, j is the round of iteration and C = 10" . Those animals whose data failed to achieve the convergence 8  criterion within the 50 iterates were edited out and were not used for the remainder of the study. Five cows failed to converge for the Von Bertalanffy and Brody's functions in Canada. These cows were edited out leaving 255 cows for further analysis. All cows in New Zealand achieved convergence for all four functions. Sensitivity of the functions was tested by using different starting values for the parameters. This was to find out whether this significantly affected the estimates of the parameters obtained at convergence. The growth functions were assessed for their goodness of fit by comparing the mean residual sums of squares (SSE) of each function, averaged over all cows in the study. The biological feasibility of the range of parameter A and its mean were also used as evaluation criteria. Heritabilities were estimated for the A, b and k parameters obtained in the New Zealand environment using the four growth functions.  The number of animals in Canada were  substantially lower than in New Zealand, consequent to which there were problems with convergence of the log likelihood function during estimation of heritabilities, particularly with the b and k parameters. No reliable estimates could therefore be obtained in Canada. The heritabilities were estimated using an animal model as follows;  y=Xb + Z a + e  (2.1)  23  where y is a vector of observations (A, b or k), b is a vector of the fixed effects of strain of sire 2  2  and herd, a ~ (0, Aa a) is a vector of additive genetic values, e ~ (0, l a e) is a vector of errors and X and Z are known design matrices of the fixed and random animal effects respectively. A is the additive genetic relationship matrix in which were included the cows, their sires, dams and maternal grandsires. I is an identity matrix. Correlations between the random effects were assumed to be zero. The model was evaluated using the DFREML program by Meyer (1991). Maximization of the log likelihood was achieved by using the simplex search procedure with convergence -8  achieved when the change in the variance of the log likelihood function was less than 10 .  2.4.3. Comparison of growth function traits  Based on the three criteria; goodness of fit, and the range and mean of parameter A, the Von Bertalanffy function was assessed to have the best fit to the data. This is discussed in detail in the results and discussion section. The function was therefore used for further comparisons of cows sired by the two sire strains for growth curve traits. The growth parameters A b and k estimated using the Von Bertalanffy function were analyzed using a mixed model in the general linear models procedure of SAS (SAS Institute Inc., 1990) to test for differences in origin (strain) of sire. The analysis was done within environment. The model used was as follows:  Y k = M + Ti + S (T ) + H Y + e ij  j  i  k  ijk  (2.2)  where;  24  Yijk - an observation on the ijkth cow \x - overall mean Ti = the fixed effect of the ith strain (sire origin) Sj(Ti) = the random effect of the jth sire within strain i HYk = the fixed effect of the kth herd-year-season(Canada) or herd (New Zealand) of birth e  ijk the random error associated with the ijkth observation =  2.4.4. Comparison of actual liveweights  Observed liveweights were compared for differences between the two strains of sire within each environment and for the effects of other fixed factors of herd and year-season of birth (Canada only). Comparisons were done for liveweights at birth, 6, 12, 18 and 30 months of age. The model used was;  Y  ijk  = M + Tj + Sj(Ti) + HY + age + e k  ijk  ijk  (2.3)  where agejjk is the covariate of age in days at the point of weighing. The rest of the terms are as shown in model 2.2. Age was fitted as a covariate to correct for actual age at weighing. The correction for age was necessary since animals were not necessarily weighed at exact ages but rather there were differences of up to 2 weeks. For 30 months, number of days post first calving was also fitted as a covariate to correct for the effect of lactation.  25  2.5. RESULTS AND DISCUSSION 2.5.1. Goodness of fit and comparisons among the curves  The overall means and standard deviations for the sums of squared errors and parameter estimates for each of the Gompertz, Brody and Von Bertalanffy growth functions are given in Tables 2.2 and 2.3 for the Canadian and New Zealand environments respectively. The sums of squared errors in New Zealand were much lower than those for Canada. This is due, at least in part, to the fact that in New Zealand, as explained in the previous chapter, the bi-monthly weights between birth and 6 months, 6 months and 18 months and 18 months and 30 months were estimated by assuming linear growth between any two consecutive points of growth. This tends to smooth the curve and thus remove the effects of environmental and physiological variations that would normally show fluctuations in the growth curve. In Canada on the other hand all the weights used were actual observed weights with all the fluctuations and, not surprisingly, the standard errors for the mean SSE are much wider. The means of sums of squared errors for the four functions in Canada given in Table 2.2 are not significantly different from each other. The Von Bertalanffy function, however, had the lowest SSE and was therefore assumed to have the best fit. The Logistic function had the poorest fit as compared to the other three functions. All cows achieved convergence at the specified maximum 50 iterates with the Gompertz and Logistic functions while the same 5 cows failed to achieve convergence with the Brody and Von Bertalanffy functions. The Von Bertalanffy and Brody functions therefore had similar results for the number of animals achieving convergence, but the former demonstrated a better fit by its measure of SSE and what was considered a more realistic estimate of the asymptotic mature weight A. The mean estimate of mature weight (A) by the Brody function (733 kg) appears to be outside of the reasonable range of the Holsteins in Canada. Working with Canadian dairy cows and four  26  Table 2.2. Overall means and standard deviations of sums of squared errors (SSE) and of growth curve parameter estimates for the three growth functions in Canada. (n=255). Growth function SSE A (kg) Gompertz  25702 (41304)  603.4 (72.0)  2.42 (0.26)  0.004 (0.0009)  Brody  26338 (40880)  733.4 (157.2)  0.975 (0.0245)  0.002 (0.0005)  Von Bertalanffy  24780 (40991)  623.0 (80.7)  0.587(0.0463)  0.003 (0.0008)  Logistic  30362 (42470)  574.9(61.7)  6.82(1.60)  0.006 (0.001)  Table 2.3. Overall means and standard deviations of sums of squared error (SSE) and of growth curve parameter estimates for the three growth functions in New Zealand. (n=677). Growth function SSE A (kg) b k Gompertz  1080 (1205)  407.4 (36.0)  2.27 (0.18)  0.004 (0.0006)  Brody  2578 (2826)  489.7 (68.4)  0.95 (0.17)  0.002 (0.0005)  Von Bertalanffy  1275 (1577)  420.8 (38.8)  0.555 (0.031)  0.004 (0.0006)  Logistic  1301 (821)  388.0 (33.3)  6.24(1.01)  0.007 (0.001)  growth functions Perotto et al. (1992) reported estimates of A ranging from 525 to 607. The difference between the Brody and the Gompertz functions in this study in their goodness of fit is consistent with the findings of Perotto et al. (1992) who however used residual mean squares from ANOVA analyses as a comparison statistic. The Brody function had higher residual mean squares than the Gompertz. They however did not try to fit the Von Bertalanffy function to their data.  27  In Canada a good maximum limit for mature weight was considered to be 850 kg. The Logistic function achieved convergence for all animals and the estimates of mature weight A ranged from 437 to 917 kg. The Von Bertalanffy function, on the other hand demonstrated the best goodness of fit and 3 cows had estimates of A exceeding 850 kg. Its estimate of mature weight ranged from a minimum of 454 to a maximum of 991 kg. The Gompertz and Brody functions had estimates of A ranging from 446 to 1209 kg and from 490 to 1630 kg respectively. The former had 1 cow with an estimate greater than 850 kg while the latter had 40 cows in that category. After comparison using all the criteria the Von Bertalanffy function was judged to be the one giving the best fit to the growth data. The mean SSEs for the four functions in the New Zealand environment were also not statistically different. All the animals in the dataset were able to achieve convergence within the maximum 50 iterates specified for all the four functions. The SSE results given in Table 2.4.2. suggest that the Gompertz function has the best fit of the four, results that are not consistent with those from the Canadian environment. This function, however, appears to under-estimate the weight at maturity (A). The mean mature weight for Holstein-Friesian cows in New Zealand is approximately 450 kg (Bevin Harris, personal communication). The Von Bertalanffy function which has the next best fit according to the SSE, gives a relatively better estimate of the mature weight with a mean of 420.8 kg. In this environment the Brody function had the poorest fit and the higher estimate for mature weight (489 kg) than the expected average of approximately 450 • kg. The Logistic function which had the poorest fit with the Canadian data fit better than the Brody, but just like in Canada tended to underestimate the mature weight A with a mean estimate of 388 kg. Sensitivity tests on the functions showed that the estimates obtained with different sets of values for the starting parameters were very similar. In most cases the values were  28  exactly the same at convergence suggesting that the estimates of the growth curve parameters were global as opposed to local. The results of the comparison of all the four functions in this study agree well with those of De Nise et al. (1985) who, working with Jersey cow data, also found that the best fit was shown by the Von Bertalanffy function when compared to the Brody and the Gompertz functions. Their results however were different when they fitted the same functions on beef cows with the ranking then being Brody, Von Bertalanffy and Gompertz. They reported that overall for both cattle types, the Von Bertalanffy model suited the data reasonably better although it overestimated weight prior to 6 months of age. The overall means for SSE across both environments were 13391, 14458, 13028 and 15832 for the Gompertz, Brody, Von Bertalanffy and Logistic functions, respectively. Von Bertalanffy model therefore appears to have the best fit of the time-weight data implying a better description of the growth curves of the study cows. The Logistic function had the tendency to underestimate mature weights in both environments and also had the worst fit in Canada. The Brody function, on the other hand, tended to give estimates higher than the expected average of asymptotic mature weight of approximately 450 kg (Bevin Harris, personal communication). This tendency by the Brody to overestimate mature weight was also reported by Lopez de Torre et al. (1992) who also recommended the Von Bertalanffy as having the best fit among Retinta beef cows. Koenen and Groen (1996) also compared several functions working with Dutch Black and White and concluded that the Von Bertalanffy function gave the best fit. The estimates of the parameter A in Canada in this study from the Gompertz, Brody and Von Bertalanffy functions are higher than those reported by Perotto et al. (1992) with the same functions, though for purebred Holsteins, Ayrshires and their crosses in Canada. The estimates  29  for b and k in both environments however do show general agreement with those from other similar studies in the literature (Brody, 1945; Laird, 1966; Koenen and Groen, 1995).  2.5.2. Heritability estimates for curve parameters  Heritabilities for the curve parameters for all four functions were estimated in the New Zealand environment and are presented in Table 2.4. The heritability estimates of the A parameter (mature weight) were similar for all four functions at .42 to .43. Similar estimates were reported by De Nise et al. (1985) after fitting Brody's and Richards functions to the growth curves of beef cows. They reported heritability estimates of .44 ± .27 for both Brody's and Richards functions.  Quaas (1973), however, found a larger heritability estimate for the A  parameter from Brody's function (.56 ± .19). Brown et al. (1972) also studied Brody's function using Hereford data and reported a heritability estimate of 34 ± .25. The estimate of heritability of A in this study for the Von Bertalanffy function is higher than the .21 ± .007 reported by Koenen and Groenen (1995) for the same function fitted on weight data of Dutch Black and White cows. The heritability estimates of the b parameter (indicates the proportion of the asymptotic mature weight to be gained after birth) across the four functions in this study are slightly more variable but in the range of .28 to .35. The .30 ± .12 estimate for the Von Bertalanffy function is similar to the estimate reported by Koenen and Groen (1995) who reported an estimate of .31 ± 0.08 with the same function while the estimate from the Brody function is similar to the .39 ± .27 reported by Brown et al. (1972). Quaas (1973) and De Nise et al. (1985) obtained much higher estimates with the Brody function of 62 ± .34 and .60 ± .19, respectively.  The heritability  estimate for k (maturation index) for the Von Bertalanffy in this study is much higher than the  30  Table 2.4. Heritability estimates of growth curve parameters in the New Zealand environment Parameter Gompertz Brody Von Bertalanffy Logistic A .43 ± .15 .43 ±.15 .43 ±.15 .42 ±.14 b  .28+ .12  .35 ±.13  .30 ±.12  .26 + .11  k  .18 ± .10  .26 ± .12  .19 ± .10  .16 + .09  0.02 ± 0.05 reported by Koenen and Groen (1995). The estimate for the Brody is similar to the .20 ± .26 estimate reported by De Nise et al. (1985) and the 20 ± .17 reported by Quaas (1973). Brown et al. (1972) reported a much higher estimate (33 ± .25), but for Hereford cattle which have been bred for production traits. It would appear from the measures of heritability that the A parameter as a trait has very similar heritabilities across all the four functions. The b and k parameters are more closely related between the Von Bertalanffy, Gompertz and Logistic models.  This similarity of  parameters across functions underlines the fact that they are all variable forms of the Richards function. In the Richards function the parameter M (inflection point) is variable whereas in the other four functions it is fixed at different constants.  2.5.3. Factors affecting growth curve parameters  As explained earlier, the Von Bertalanffy model was adopted as the model that best described the growth curves of project cows based on the a lower mean SSE, the number of animals achieving convergence and a more realistic mean mature weight (A). The curve parameters obtained by fitting the Von Bertalanffy function were then further analyzed to test for  31  differences among the strains and the effects of other fixed variables within each environment. The results of the analysis of variance for the parameter estimates from the Von Bertalanffy function in Canada are presented in Tables 2.5 while the least squares means and standard errors for the curve parameters from the same function are given in Table 2.6. The corresponding results for New Zealand are presented in Tables 2.7 and 2.8, respectively. In Canada the effect of strain of sire was significant for mature weight (A) (P < 0.05). The difference between cows sired by Canadian bulls and those by New Zealand bulls in mean mature weight estimated by this function was approximately 22 kg. The effects of sire within strain of sire and herd-year-season of birth were also significant for mature weight. For the parameter b, strain of sire was not an important source of variation. The lack of significant difference implies that the proportions of mature weights at birth for the cows sired by both strains were relatively similar, even though the two groups grow to significantly different mature weights. This lack of variability in the b parameter of the Von Bertalanffy function across different breeding groups was also observed by Brown et al. (1976) who associated this with the magnitude and non-variable nature of the exponent M which in this function is assigned a value of = 3. M determines the proportion of the asymptotic mature weight at which the growth inflection point occurs. Herd-year-season of birth was significant for the b parameter, but sire within strain of sire was not. Effects of strain of sire and sire within strain of sire were both not significant for parameter k (maturation index). The implication of a similar parameter k is that in Canada cows sired by bulls of either strain tended to reach their asymptotic mature weight at approximately the same age. The effect of herd-year-season of birth was a highly significant (P < 0.01) source of variation..  32  Table 2.5. Analysis of variance (R2 x 100) results for the Von Bertalanffy function parameters in Canada Effect df A b k Strain of sire  1  1.4*  1.0  0.7  Sire (Strain)  38  17.4*  7.2  8.9  HYS of birth  32  21.7*  35.3*  23.0  Model  71  41.7*  47.6*  34.0  Error  180 10.8  6.5  22.0  CV * significant at P < 0.05 HYS = herd-year-season of birth  Table 2.6. Least squares means and standard errors for growth curve parameters estimated using the Von Bertalanffy function for cows in Canada Strain Canadian (n = 138) New Zealand (n = 114) A  637* (8.7)  615 (9.3)  b  0.595 (0.005)  0.606 (0.005)  k  0.0032 (0.0001)  0.0033 (0.0001)  * means significantly different at P < 0.05  Table 2.7. Analysis of variance (R x 100) results for the Von Bertalanffy function parameters in New Zealand Effect df A b k Strain of sire  1  2.1*  0.1  0.02  Sire (Strain)  38  13.1*  5.4*  4.7*  Herd  18  14.4*  44.9*  48.3*  Model  57  33.3  62.8*  66.3  Error  619 8.0  3.6  9.5  CV * significant at P < 0.05 HYS = herd-year-season of birth  Table 2.8. Least squares means and standard errors for growth curve parameters estimated using the Von Bertalanffy function for cows in New Zealand Strain Canadian (n = 138) New Zealand (n = 114) A  429* (1.9)  416 (2.0)  b  0.556 (0.0012)  0.559 (0.0012)  k  0.0037 (0.00002)  0.0037 (0.00002)  * means significantly different at P < 0.05  In New Zealand (Table 2.7) asymptotic mature weight (A) was significantly (P < 0.05) affected by the strain of sire. The mean mature weights (Table 2.8) were 429 kg and 416 kg for cows sired by the Canadian and New Zealand bulls, respectively. These weights however appear to be slight under-estimates compared to the observed mature weight of New Zealand Holstein Friesian cows of approximately 450 kg (Harris, B. L., personal communication). The mean weight for the cows sired by Canadian bulls would then be expected to be higher than that as this study has shown that daughters of Canadian bulls are slightly heavier than those of New Zealand bulls. The difference between average weight at 30 months predicted using the function and the average of the actual (observed) weight at the same age were not significantly different, implying that the estimates by the function up to that age were quite accurate. It is only the mean asymptotic weight that appears to be slightly underestimated in this case. This was probably due to the lack of data at older ages and what the function really gives is an extrapolation. This should however not create any problem with the comparison of sires since the under-estimation occurs in both strains and there is no indication that the relative magnitude of the underestimation is different in the two strains. The effects of sire within strain of sire and herd were both important (P < 0.05) sources of variation for estimated mature weight (A). This implied that within strains sires had daughters that differed significantly for mature weight and this might have implications later when feed requirements are computed and hence their profitabilities. For the b and k parameters strain of sire was not significant but sire within strain of sire and herd were significant. Like in Canada the lack of a significant difference in the b parameter between strains implies that the proportions of mature weights at birth for the cows sired by both strains were relatively similar, even though the two groups weigh significantly different at maturity. Likewise a similar parameter k implies comparable rates of maturing for the two strains.  35  The fact that the sire within strain was significant for all the three parameters in the New Zealand environment while it was not in the Canadian environment might suggest the existence of a genotype by environment interaction at the micro level. This means that sires within a given strain perform differently in the two environments in that the magnitudes of differences between them are bigger in one environment than the other. The mean growth curves estimated by the Von Bertalanffy function in each of the environments are shown in Figure 2.1. First each individual cow's monthly live weights were estimated using the Von Bertalanffy function. The strain "mean" curves as shown in the Figure 2.1 were then obtained by computing the average weight every month for cows within strain of sire.  2.5.4. Comparison of observed live weights  Observed liveweights were compared for the heifers sired by the two strains of sires in both environments. The animals were compared for weights at birth, 6 months 12 months (Canada only), 18 months and at 30 months which was approximately six months after first calving. The ANOVA table for the Canadian environment is shown in Table 2.9 and the means are given in Table 2.10. The corresponding results for the New Zealand environment are given in Tables 2.11 and 2.12 respectively. In Canada there was no significant differences between the heifers sired by the two strains in weights for all ages from birth to 30 months. The random effect of sire within strain and the fixed effect of herd-year-season of birth were both important for weight at all ages except at 18 months when sire within strain was not important. At birth the mean liveweights were similar for daughters of both sire strains at approximately 43 kg. At 6 months calves sired  36  Figure 2.1. Mean Growth curves fitted by the Von Bertalanffy functionfor cows sired by Canadian and New Zealand bulls in the Canadian and New Zealand environments. 700 ,  0  _,  300  600  900  1200  1500  1800  2100  Age (days)  CN/CN NZ/CN CN/NZ NZ/NZ  = Canadian sired cows in the Canadian environment = New Zealand sired cows in the Canadian environment = Canadian sired cows in the New Zealand environment = New Zealand sired cows in the New Zealand environment  37  Table 2.9. Analysis of variance results (R x 100) for heifer liveweights at birth, 6 12, 18 and 30 months of age in Canada. Age (months) 2  Effect  df  Birth  6  12  18  30  Strain of sire  1  0.04  0.4  0.01  0.01  1.5  Sire (Strain)  38  23.5*  14.0  15.7*  14.7  20.9  HYS  35  16.9*  39.1*  52.0*  34.2*  22.9*  43.6*  59.3*  70.6*  50.0*  49.7*  12.6  9.8  8.5  8.9  12.6  Model Error CV  198  * effect significant at P < 0.05 HYS = herd-year-season of birth  Table 2.10. Least squares means and standard errors of liveweights for heifers sired by Canadian and New Zealand bulls in Canada . Strain of sire 1  Age (months)  1  Canadian (n = 144)  New Zealand (n = 114)  Birth  43.1 0.7  42.8 0.8  6  192.1 2.5  188.1 3.0  12  346.2 4.3  347.3 4.6  18  444.2 5.9  443.1 6.3  30  564.6 10.8  542.7 10.7  Mean weights at all ages not significantly different (P > 0.05)  38  by Canadian bulls were 4 kgs heavier and by 30 months of age they weighed approximately 22kgs more than their counterparts sired by New Zealand bulls. All these differences were however not significant, possibly due to the relatively small numbers of animals in the study hence bigger standard errors. A study of the same animals by Peterson (1991) showed that in both Canada and New Zealand heifers sired by Canadian bulls were also significantly taller at the withers than those sired by New Zealand bulls at birth, six, and eighteen months of age. In New Zealand (Tables 2.11 and 2.12), contrary to the results in Canada, the two strains of sire were significantly different in daughter liveweights at the four ages considered. Both sire within strain and herd were also important for weight at all ages. Heifers sired by Canadian sires were consistently heavier from birth to 30 months of age. The difference increased from 1.5 kg at birth to approximately 12 kg at 30 months of age. Thus even though the differences between strains in New Zealand were smaller than those in Canada, they were significant, possibly due to more records, the relatively lower liveweights at the given ages, and the relationship between the mean and the standard deviation. The overall mean birth weight in Canada was 43.7 kg while that in New Zealand was 36.9 kg. The corresponding means at 6, 18 and 30 months were 190.1 vs. 160.0, 443.5 vs. 351.0 and 553.7 vs. 383.6 kg respectively. The significance of the differences in New Zealand might imply that the differences seen are due to nutritional and environmental constraints, under which circumstances the animals are not able to gain weight as much as their genetic potential would allow. In Canada on the other hand, nutrition might not be a constraint to weight gain and thus daughters from both strains are able to express more of their genetic potential. For the ages tested and the data available, this potential seems to be quite similar for both strains of bulls. Possibly a larger number of observations (animals) might reveal differences as were realized in New Zealand.  39  Table 2.11. Analysis of variance results (R x 100) for heifer liveweights at birth, 6, 18 and 30 months of age in New Zealand. Age (months) 2  Effect  df  Birth  6  18  30  Strain of sire  1  1.8*  1.1*  1.4*  2.4*  Sire (Strain)  38  9.5*  6.1*  7.1*  13.6*  HYS  18  17.3*  41.1*  43.2*  13.4*  32.8*  65.1*  62.0*  33.2*  10.6  8.8  7.2  7.5  Model Error  620  CV effect significant at P < 0.05  Table 2.12. Least squares means and standard errors of liveweights for heifers sired by Canadian and New Zealand bulls in New Zealand . Strain of sire 1  Age (months)  Canadian (n = 144)  New Zealand (n = 114)  Birth  37.7 (0.2)  36.1 (0.3)  6  158.8(0.8)  153.5 (1.1)  18  356.4(1.5)  345.6 (2.0)  30  390.4(1.7)  377.0 (2.3)  Means at all ages significantly different at P < 0.05  The effect of herd-year-season (herd in New Zealand since all calves were born in the same season and year) of birth was an important source of variation for weights at all ages (P < 0.05). In New Zealand sire within strain of sire was also significant (P < 0.05) for weights at all ages. In Canada sire within strain of sire was important only for liveweight at birth and at 12 months but not at 6, 18 and 30 months of age. When weights at the same ages but estimated using the Von Bertalanffy growth function are subjected to the same analysis, the differences between sires within strain in New Zealand are consistent with those seen in observed weights. In Canada, with both observed and estimated weights, sire within strain was significant at 12 months and non-significant at 30 months. There were slight differences at 6 and 18 months of age where no significance of sire within strain was observed with the estimated weights whereas observed weights showed significant differences. The effect of sires within strain in Canada was important for both observed and estimated mature weights. This consistency between results from live weights and those predicted using the growth function suggests that the Von Bertalanffy growth function gave realistic estimates.  2.6.  CONCLUSIONS  Of the four growth functions tested the Von Bertalanffy had the lowest mean SSE across both environments. It also gave estimates of mature weights that were considered close to the expected mature weights for the populations. The Brody function tended to overestimate the mature weight and had the largest estimates of SSE in both the Canadian and New Zealand environments. In New Zealand, the Gompertz function, had the lowest estimate of SSE, but  41  tended to seriously underestimate the asymptotic mature weight A. However, the Von Bertalanffy function apparently slightly underestimated the A parameter in this environment. When all the criteria were considered it was decided that the single growth function of choice to describe the growth curves of the cows in both environments is the Von Bertalanffy function. The heritability estimates for the three parameters in the New Zealand environment are in the medium to high range and are similar to those reported in the literature. The parameter A was uniformly heritable across all the three functions probably due to the mathematical similarity of the growth functions themselves. The heritabilities of the parameters b and k were slightly more variable among the three functions, implying small differences in the interpretation of the traits they represent in each of the functions. In Canada, strain of sire significantly affected mature weight A. Cows bom of Canadian sires were on average 22 kg heavier at maturity than their counterparts sired by New Zealand sires. Strain of sire, however, was not important for the maturation index (parameter k), implying that daughters of either strain attained their asymptotic mature weight at approximately the same age. The proportion of mature weight at birth, represented by parameter b also did not show any significant difference across strains of sire. The effects of sire within strain of sire was important for weight at maturity but not important for parameters b and k in Canada. The conclusion can then be made that daughters of Canadian bulls in Canada tend to mature heavier than those of New Zealand bulls but that the general nature of their growth curves is not very different in this environment. In the New Zealand environment daughters of Canadian bulls were significantly heavier at maturity. Differences were also significant for all three parameters among sires within strains, implying that sires do differ in the growth traits of their progeny. Strain of sire, however, was not important for the proportion of mature weight at birth, b, and the rate of maturing, k. Thus  42  like in Canada although the two strains grow to different mature weights, they matured at the same rate. Comparing the observed liveweights from birth to 30 months, the two strains showed significant differences for all ages in New Zealand but were not significantly different for any of the ages in Canada. It can be concluded that the New Zealand sired heifers are able to express more of their genetic potential for weight gain in the Canadian environment than in their native one. Differences in liveweight also mean differences in feed maintenance requirements. Mature weights in both environments have been seen here to be different between daughters of the two strains of bulls. What this study will further seek to test is whether these differences in liveweight, coupled with any differences that might exist in production, will necessarily result in important differences in economic efficiency. A question also does arise whether an important genotype by environment interaction would be realized in economic efficiency, as well as in weights and production traits.  43  2.7.  REFERENCES  Bakker H. and Koops, W. J., 1978. An approach to the comparison of growth curves of Dutch Friesian, British Friesian and Holstein Friesian cows. Patterns of growth and development in Cattle: A Seminar in the EEC Programme of Coordination of Research on Beef Production, Ghent, October, 1977. Current Topics in Veterinary Medicine. The Hague, Martinus Nirjhoff, 1978 v2. pp 705-715 Barbato, G. F., 1991. Genetic architecture of growth curve parameters in chickens. Theor. Appl. Genet. 83:24-32. Brinks J. S., Clark, R. T., Keiffer, N. M. and Urick, J. J., 1964. Estimates of genetic, environmental and phenotypic parameters in range Hereford females. J. Anim Sci. 23: 711-716. Brody S., 1945. Bioenergetics and growth. Reinhold Publ. Co., New York, NY. Brown, J. E., Brown, C. J. and Butts, W. T., 1972. A discussion of the genetic aspects of weight, mature weight and rate of maturing in Hereford and Angus cattle. J. Anim. Sci. 34: 525537. Brown, J. E., Fitzhugh Jr., H. A., and Cartwright, T. C , 1976. A comparison of nonlinear models for describing weight-age relationships in cattle. J. Anim Sci. 42: 810-818. DeNise, R. S. K. and Brinks, J. S., 1985. Genetic and environmental aspects of the growth curve parameters in beef cows. J. Anim Sci. 61: 1431-1440. Doren P. E., Baker, J. F., Long, C. R. and Cartwright, T. C , 1989. Estimating parameters of growth curves of bulls. J. Anim Sci. 67: 1432-1445. Eisen, E. J., Lang, B. J. and Legates, J. E., 1969. Comparisons of growth functions within and between lines of mice selected for large and small body weight. Theoret. Appl. Genetics 39:251-260. Finney, D. J., 1978. Growth curves: Their nature uses and estimation. Patterns of growth and development in Cattle: A Seminar in the EEC Programme of Coordination of Research on Beef Production, Ghent, October, 1977. Current Topics in Veterinary Medicine. The Hague, Martinus Nirjhoff, 1978 v2. pp 658-672. Fitzhugh, H. A., Jr., 1976. Analysis of growth curves and strategies for altering their shape. J. Anim. Sci. 42: 1036-1051. Koenen, E. P. C. and Groen, A. F., 1995. Genetic analysis of growth patterns of black and white dairy heifers. J. Dairy Sci., 79: 495-501.  44  Laird, A. K., 1966. Postnatal growth of birds and mammals. Growth 30: 349-363. Cited in Perroto, D., Cue, R. I. and Lee, A. J., 1992. Comparison of nonlinear functions for describing the growth curve of three genotypes of dairy cattle. Can. J. Anim. Sci. 72: 773782. Lopez de Torre, G., Candotti, J. J., Reverter, A., Bellido, M. M., Vasco, P., Garcia, L. J. and Brinks, J. S., 1992. Effects of growth curve parameters on growth efficiency. J. Anim. Sci. 70: 2668-2672. Meyer, K., 1991. Estimating variances and covariances for multivariate Animal Models by Restricted Maximum Likelihood. Genet., Sele., Evol. 23: 67-83. Pearl, R. and Reed, L. J., 1923. On the mathematical theory of population growth. Metron 3: 619. Perotto, D., Cue, R. I. and Lee, A. J. 1992. Comparison of nonlinear functions for describing the growth curve of three genotypes of dairy cattle. Can. J. Anim. Sci. 72: 773-782. Peterson R., 1991. Evidence of a genotype/environment interaction between Canadian Holstein and New Zealand Friesian cattle under Canadian and New Zealand management systems. Proc. of the 42nd Annual Meeting of the European Association for Animal Production, Berlin, Sept. 9-12, Vol. 1: pp 49. Quaas, R. L., 1973. Genetic variation in growth curves of Hereford females. PhD. Dissertation. Colorado State Univ., Ft. Collins. Cited in DeNise, R. S. K. and J. S. Brinks. 1985. Genetic and environmental aspects of the growth curve parameters in beef cows. J. Anim Sci. 61: 1431-1440. Ralston, M. L. and Jennrich, R. I., 1978. DUD, A derivative-free algorithm for nonlinear least squares. Technometrics. 1: 7-14. Richards F. J., 1959. A flexible growth function for empirical use. J. Exp. Botany, 10: 290-300. Statistical Analysis System (SAS) Institute, Inc., 1994. SAS Users Guide, Release 6.08. SAS Institute Inc., Cary, N. C. Timon, W. M. and Eisen, E. J., 1969. Comparison on growth curves of mice selected and unselected for postweaning gain. Theor. Appl. Genet. 39: 345. von Bertalanffy, L., 1957. Quantitative laws in metabolism and growth. Q. Rev. Biol. 32: 217231. Winsor, C. P., 1932. The Gompertz curve as a growth curve. Proc. Naty. Acad. Sci. U.S.A. 18: 1-8.  45  CHAPTER 3: GENETIC ANALYSIS OF ECONOMIC EFFICIENCY OF DAUGHTERS OF CANADIAN AND NEW ZEALAND SIRES IN CANADIAN AND NEW ZEALAND DAIRY HERDS  3.1. SUMMARY  A measure of economic efficiency was derived for daughters of Canadian and New Zealand sires in Canadian and New Zealand herds from body weights estimated using the Von Bertalanffy growth function, milk production, calving records and milk prices used in both countries.  The measure of economic efficiency, PF , was defined as (milk revenue)/(feed ;  requirements) up to the end of lactation i, or up to the end of the cow's productive life. Milk revenue was calculated from milk prices in both countries with two systems of pricing in Canada that are used in British Columbia and Ontario. Heritabilities for economic efficiency traits and first lactation production traits were estimated in both environments and genetic and phenotypic correlations between these were also estimated for New Zealand data. Heritability estimates for economic efficiency traits in Canada were low to moderate ranging from 0.12 (economic efficiency up to the end of lactation one with Ontario prices) to 0.21 (lifetime economic efficiency with Ontario prices). The estimates were a little higher in New Zealand at 0.21 (economic efficiency up to the end of lactation one) and 0.34 (lifetime economic efficiency). Genetic correlations between first lactation yield traits and economic efficiency traits in New Zealand were medium to high ranging from 0.44 (milk yield and lifetime economic efficiency to 0.70 (fat/protein yield and economic efficiency up to the end of lactation one). Genetic correlation between economic efficiency to the end of lactation one and lifetime was 0.73  46  In New Zealand daughters of New Zealand sires were more efficient over productive life than daughters of Canadian sires. Strain differences were not significant for any other economic efficiency traits in either environment.  3.2. INTRODUCTION  The utility of a dairy cow to her owner depends on profitable performances across a relatively long productive life. The challenge to producers and breeders is to select for the combination of traits most conducive to profitable lifetime performance and to provide the management conditions which facilitate optimum expression of the traits selected (Cassell and Weigel, 1994). The profitability of a cow depends on the combination of a number of factors that are related to both the performance of the cow and herd management. Other specific nonproduction traits such as age at first calving, days open, days dry and length of productive life are important factors influencing the economic efficiency/profitability of a cow and are partly determined by quality of management. In order to understand the interrelationships of the cow and management factors involved in the economic performance of a dairy cow/enterprise, it is necessary to carry out research involving detailed collection of the relevant information. In most circumstances however, it is prohibitively expensive to collect this data over a lifetime of a large number of cows in deliberately fashioned experiments. This limits the efforts of research to generate accurate measures of economic performance using the data collected in conjunction with the usual milk recording programs, for example the Dairy Herd Improvement (DHI) programs in  47  Canada and in the United States. Tigges et al. (1984) conducted a study to evaluate the accuracy and applicability of the results from studies involving the use of these data. They examined the accuracy of relative net income of individual cows derived using field data compared to a thorough economic analysis involving the collection of detailed economic and production information. In their study they used detailed data on birth and freshening weights, weight gains, milk yield, weight at culling, number of cases of mastitis and number of breedings. They concluded that relative net income and relative net income per day of productive life derived from field data were accurate estimates of relative cow profitability. This means that results from appropriate field (for example DHI) data can give credible and unbiased estimates of production circumstances. The dataset used in this study (Canadian New Zealand genotype by environment interaction trial - CANZ) is intermediate between a small detailed dataset and general field data. This is because unlike most field data, weight records were available that can be used for more specific analyses. The objective of the trial was to test for the existence of a genotype by environment interaction for production traits in dairy cattle (Peterson, 1991). The focus of the study was on Holstein-Friesian strains from Canada and New Zealand because management and feeding programs of the dairy herds in the two countries are very different. Secondly, selection programs for increased production in the two countries may have caused the populations to diverge for traits associated with adaptation to the differences in environmental conditions in the two countries (Peterson, 1991). One of the ways to go about improving lifetime economic efficiency of dairy cows is by analyzing cow efficiency itself, its relationship with other traits of the cow and then combining these evaluations with other available genetic evaluations. The proposed study will try to gain further understanding of such relationships through the following objectives;  48  (i) To determine and compare lifetime and partial lifetime economic efficiency of daughters of Canadian and New Zealand sires in both countries. In Canada two pricing systems, fluid and component, will be considered. (ii) Examine the relationship between measures of lifetime and partial lifetime economic efficiency of dairy cows. (iii) To determine the relationship between the measure of economic efficiency and other cow traits, especially those measured early in life and thus determine the potential of using such traits in predicting sire economic efficiency. The traits whose relationship with economic efficiency will be investigated are, first lactation milk, fat and protein yield, second lactation milk, fat and protein yield, length of productive life, and number of calvings.  3.3. REVIEW OF RESEARCH ON PROFIT FUNCTIONS IN DAIRY CATTLE  Measures of profitability for either a single animal or a defined population are phenotypic measures which contain both genetic and environmental factors leading to differences in economic performance. The aim of research with profit measures or functions is partly to identify the economically important traits and estimate the relative importance of these. The traits can then be combined in a selection index to achieve an economic breeding goal. Most of the studies done (Harris, 1970; Andrus and McGilliard, 1975; Gilmore, 1977; Balaine et al., 1981; Norman et al, 1981; Bertrand et al, 1985; de Haan et al, 1992; Weigel, 1993; MacAllister et al, 1994; Visccher and Goddard, 1995) differ in the methods of estimation of the  49  profit trait(s) and in the number of variables examined. The studies have also differed in their depth of detail and size. Some of the studies were small and complete (Andrus and McGilliard, 1975; Gilmore, 1977; Balaine et al, 1981 and Bertrand et al., 1985). They were conducted utilizing detailed economic records and other production parameters but were small in scale since they involved few cows and single or at most few herds. These studies have the advantage of providing detailed information that is not usually collected under normal dairying conditions. However, since management has a large effect on lifetime net income (Pearson and Miller, 1981; Rogers et al., 1988), when such studies are geared towards the identification of which traits are of general economic importance, a larger cross-section of systems (herds) would be desirable. Apart from the studies done on a small scale, others have been done on larger and more diversified data sets. These studies are termed "big" since they involved the use of large volumes of field data collected by the milk recording associations. Such data sets are usually not very detailed in terms of economic information, in that unlike in the small studies cited above, no records are available for variables like feed intake, liveweight, disease incidences, veterinary costs and so on. Such variables are usually left out of the derivations of profit functions or else extracted from literature. Examples of large studies include Beaudry et al. (1988), Cassell et al. (1990), Ruiz (1991) and de Haan et al. (1992). In their study for example, de Haan et al. (1992) worked with records of over 70,000 Holstein-Friesians. The small studies have proved useful in the validation of bigger studies involving the use of field data (Tigges et al., 1984). As stated in the previous section, the current study can be considered intermediate between large and small in that the amount of data is not very large, but it covers many herds across two environments which are different in terms of production methods and levels, milk pricing, and marketing programs. This dataset also provides individual  50  animal weight records which are not usually available in field data. The weight information make it possible to estimate feed requirements and hence costs more realistically than if no records were available at all.  3.3.1. Different measures of profitability  Various researchers have developed different measures of profitability for use as a breeding goal for selection programs. These include lifetime net income (Norman et al., 1981), income per day of productive life (Harris, 1970; Weigel, 1993), net income adjusted for opportunity cost (Van Arendonk, 1991; de Haan et al., 1992; Weigel, 1993), discounted net income (MacAllister et al., 1994) and income/expense ratio (Harris, 1970). Net income (also referred to as relative net income) is defined as the returns from the sales of milk, calves and cull cows less variable costs such as feed and labour and fixed costs such as building costs. The major difference in the studies using this function appear to be in the method of estimating the various costs and expenses and in the exhaustiveness of the variables included in the two factors. Pearson and Miller (1981), outlined the following composite income and expense variables for the net income model used to estimate economic performance for different time scales;  Income =  (milk produced - milk discarded) x price adjusted for composition + calf value + ending inventory value (depending on weight of cows sold for beef or on expected future profitability for cows still in the herd or sold for dairy)  51  Expenses =  initial inventory value (dependent on weight, age atfirstcalving, and possibly on the value (based on pedigree) of the animal at birth) + feed costs (actual feed costs, or Natiaonal Research Council (NRC) requirements for maintenance, growth, production and reproduction; or actual concentrate + forage (adjusted concentrate intake and the weight of the cow)) + facilities cost (constant) x days + labour costs (constant) x days + deviation in labour for individual cows care (no. of occurrences x price) + veterinary cost (no. of occurrences x price for a type of treatment) + insemination cost (no. of inseminations, adjusted for conception rate of the bull x cost)  With field data many of the above variables are not available or are only partially available and different studies are structured and handled in accordance with the information available on the various variables. In some situations the feed consumed has not been recorded and as suggested by Pearson and Miller (1981), feed requirements computed using NRC standards could be used in the computation of the feed costs. In the CANZ trial feed intake information was not available and in the current study nutrient intake can only be estimated by the use of standards. This has previously been done by Dunklee et al. (1994) who developed a linear profit function based on the works of Pearson and Miller (1981) and Bertrand et al. (1985). Since actual feed consumption was not recorded, the researchers estimated nutrient requirements to calculate feed costs.  52  Net income per day of productive life is another form of net income that has been used in the economic analysis of dairy cattle production data (Balaine et al., 1981; Pearson and Miller, 1981). Net income per day of life, however, has non-linear properties over the range of possible length of productive life (Weigel, 1994). The non-linearity of the function occurs for cows with a very short productive life, such that a cow with only one year of productive life appears to be twice as bad in terms of production as the cow that produces for only two years. Balaine et al. (1981) found that profit per day had stronger correlations than total profit with other measures of efficiency. They concluded that profit per day was a better measure of profitability because of its strong relationship to economic efficiency and because it is more understood by dairymen. Net income, however, has been criticized by Van Arendonk (1991) who argued that net income per day of life is not applicable to dairy cattle since there are differences among cows for length of productive life. Van Arendonk therefore proposed adjusting net income functions for the opportunity cost of postponed replacement. Opportunity cost discounts returns from a cow with the assumption that she could have been culled at any time and replaced with a heifer. This use of the adjustment of the cows profit was adopted and used by Cassell et al. (1992) and de Haan et al. (1992) in their studies of the relationships of profit functions and other biological and environmental factors. The adjustment for opportunity cost reflects the replacement of culled cows by an average cow for the herd rather than a cow of equal economic value to the culled cow (de Haan et al., 1992). The opportunity cost therefore reflects the average profit of cows freshening for the first time in the same herd and year as the cow in question. Another criticism put forth against net income and net income adjusted for opportunity cost is that they do not discriminate between cows with different rates of income. The returns and costs realized at a given time are worth more than those received at a future date. This can be adjusted by the use of the concept of discounting income and expenditure to a constant time  53  and is already being used in the economic analysis of animal breeding data (Gilmore, 1977; Stott and DeLorenzo, 1988; Cassell et al, 1992; MacAllister et al, 1994). Since the discounting is done in recognition of the fact that money now is worth more in the future, the discount factor used is usually equivalent to the interest rate. Smith (1978) recommended that since interest rates are made higher by inflation, the interest rate used should be inflation-free. Harris (1970) studied a number of profit measures, one of which was income/expense ratio as a measure of economic efficiency. This form of profit function has a practical flexibility in that the numerator and denominator in the ratio need not be in the same units. This is of importance in situations where some of the variables necessary to calculate net income are not measured and can only be estimated via indirect measures. Such a scenario is seen in the study by Visscher and Goddard (1995) in which they studied the profitability of Australian dairy cattle. They studied field recorded production data from a grazing system where amount of feed consumed was not recorded. They used estimates of food requirement in terms of energy as their measure of cost and their profit trait then was income/unit of food required. Arguments can be advanced for and against any of the measures of profitability and economic efficiency mentioned here, but the use of any one of them depends on the circumstances and the availability of relevant information. Relationships among some of these measures have been examined (Balaine et al., 1981; Cassell et al, 1992; Jairath et al, 1995) with variable results. The general trend is that most of the profit functions are highly and positively correlated. Jairath et al. (1995) found phenotypic and genetic correlation estimates of .63 and .92 between profit in the first lactation and profit per day of first lactation, respectively. Cassell et al. (1992) estimated phenotypic correlations between three different measures of net income, relative net income, discounted relative net income and relative net income adjusted for  54  opportunity cost among American Holsteins. The estimates were high and positive ranging from .96 to 1.0. Balaine et al. (1981) used four functions as measures of profitability; total profit (income - expense), profit per day, profit per unit of investment (income/expense) and cost per unit of production (expense/income). They found highly positive product moment correlations between the first three measures (> .75) and highly negative correlations between the fourth (expense/ income) and the rest. This is not surprising since expense/income is simply an inverse of income/expense. Profit per day had stronger correlations than total profit with other measures of efficiency. In the current study use of any one of the functions will depend on the suitability of information available and information on feed intake, health matters, calf birth weights, number of services per conception and so on was not available. Thus any profit measure that involves net profit would not be possible to work with. An income/expense ratio in the form of that used by Visscher and Goddard (1995) seems more fitting and it is referred to as economic efficiency. This gives a fair indication of profit since efficiency and profit can be shown to be linearly related as follows;  Pr ofit = Revenue - Feed Costs - Other Costs  = Feed Costs]  Revenue  Other Costs  Feed Costs  Feed Costs  1 = Feed Costs] Pr ice of Feed  Other Costs Re venue -1 Feed Re quirements Feed Costs  Other reasons justifying the use of this economic efficiency ratio as an indicator of profitability are dealt with later in the next chapter when the ratio itself is actually computed.  55  3.3.2. Identification of economically important traits  Recently there has been an increased push among breeders to channel selection efforts towards improved lifetime performances of cows (Blake, 1984; Tigges et al., 1984; Beaudry et al., 1988; Jairath et al, 1995). Traits which are strongly related genetically to lifetime profit and measured early in life could be used to improve lifetime performance traits for which direct selection is not very practical. Several studies (Everett et al., 1976; Van Doormal et al., 1986; Dentine et al., 1987; Cassell et al., 1990; Ruiz, 1991; Visscher and Goddard, 1995) have studied these relationships using sire models while (Rogers et al., 1991,) did it using the animal model. Returns from milk production contribute the majority of the income generated from a dairy cow. Its relationship with the overall profit is therefore the most important one as far as the industry is concerned. Of utmost importance would be the relationship between the milk yield over the first one or two lactations to the lifetime profitability of the cow. Reports from the literature show on average, high and positive correlations between first lactation milk yield and profitability over various described periods of production (Stott and deLorenzo, 1988; de Haan et al, 1992; Jairath et al, 1995; Visscher and Goddard, 1995). Stott and deLorenzo (1988) reported a partial phenotypic correlation of .58 between milk yield and profit over one lactation. Jairath et al. (1995) reported a phenotypic correlation estimates of .96, .97 and .96 between milk, fat and protein yield respectively, and profit over the first lactation. The corresponding genetic correlations were also .96, .97 and .96. These were close to the results of Visscher and Goddard (1995) who reported an average genetic correlation of .79 ± .06 and .93 + .03 between profit over one lactation and the yields of milk, milk fat and protein in Holsteins and Jerseys respectively. The study of Visscher and Goddard used income/expense ratio as the measure of profit while the Jairath et al. used a net income function. Cassell et al. (1993) also reported phenotypic correlations between milk yield and different measures of profit in the range .88 to .96.  56  A farmers goal is to have a cow with a long profitable herd life. Even though literature indicates the correlation between milk yield and profitability over a given production period is positive, the use of lifetime milk records to assess the lifetime genetic merit for either sires or dams would slow down the selection process by extending the length of generation intervals. The selection of bulls used in Artificial Insemination (Al) has mostly been based on first lactation records of their daughters mostly based on the justification that first lactation yield has high genetic correlations with yields from later lactations (Meyer, 1983; Meyer, 1985; Pearson et al, 1986.). Positive genetic correlations have been reported between first lactation yield traits and measures of lifetime performance (Gill and Allaire, 1976; Lin and Allaire, 1977 and 1978; Gilmore, 1977; Hoque and Hodges, 1981; Norman et al., 1981). The accuracy and precision of these estimates has however been questioned since they were obtained using traditional ANOVA methods. Jairath et al. (1995) argue that ANOVA estimators assume random sampled data and therefore have a tendency to give biased estimates in the presence of selection. They also argue that in the majority of these studies, relationships were obtained through univariate procedures and were therefore not free of the effects of selection on correlated traits. Visscher and Goddard (1995) used REML (Restricted Maximum Likelihood) methodology with a sire model and reported genetic correlations of 0.77, 0.87 , 0.89 between lifetime profit and first lactation milk, fat and protein yield respectively among Australian Holsteins. The results they found among Jerseys were very similar to those among the Holsteins. The phenotypic correlations were in the range of 0.57 - 0.60. The correlations between first lactation milk yield and profit over a lifetime were very similar to the correlations between first lactation yield and first lactation profit. This could possibly be explained by the fact that the profit measure in their study was a ratio of income and energy requirements, which essentially could be viewed as a sort of economic efficiency. The results therefore showed that the  57  economic efficiency over a lifetime and over a single lactation are similar traits. Jairath et al. (1995) using an animal model reported phenotypic and genetic correlations of .56 and .88 between first lactation milk yield and lifetime profit. The correlations with fat and protein yield were similar to those with milk. In their study they found that even though profit traits were highly and positively correlated with fat and protein yields, they were negatively correlated with the percentages of these in milk. This possibly had something to do with the milk pricing system used in the study. Cassell et al. (1993) also found high phenotypic correlations between first lactation milk yield and profit measures up to different ages of production. They reported correlations between .53 and .63. The correlations between the yield traits and profit measures will essentially relate to the milk pricing system, that is component versus fluid. In markets where components have higher weights the correlations between component yields and profit will be high and relatively lower in fluid markets. Reproductive traits such as age at first calving (Cassell et al, 1993) and calving interval (Stott and deLorenzo, 1988) have also been reported to influence the profitability of cows. Cows with long calving intervals will tend to have bigger overhead and feed costs mostly due to longer dry periods, and consequently reduced profits per lactation. Stott and deLorenzo (1988) reported a partial phenotypic correlation estimate of -.26 between calving interval and profit. They found that calving interval has a close association with higher feed and dry period costs, and also that its correlation with overhead costs was unity. Cassell et al. (1993) reported that age at first calving was negatively correlated with net income over different herd life opportunities ranging from 48 to 72 months, implying the younger the heifer at freshening the higher the profit. They, however, found that these correlations decreased as the cow's herd life opportunity increased. This could possibly be explained by the fact that, the younger the animal, the higher the amount of energy channeled  58  towards growth and at the same time the production is relatively low only peaking at the fourth or fifth lactations. Relatively then, the benefit cost ratio would be lower over a shorter production life. In a study of 933 Holstein cows in Ohio institutional herds, Gill and Allaire (1976) also reported that age at first calving had strong genetic correlation's with lifetime net income. Their results agree with those of Lin and Allaire (1977, 1978) in a study of 1806 Holstein cows in 38 herds in California. Length of productive life has been reported to be a major factor in cow profitability because economic returns of a cow depend partly on her parity (Renkema and Stelwagen, 1979; Hocking et al., 1988). That is, the cumulative economic returns will usually increase with increase in parity and this has direct implications on the decision of whether or not to cull a cow. The question of when to replace a cow is an economic one, and the decision to not replace means she is likely more profitable than the potential replacement (Ruiz, 1991). Length of productive life has been reported (Andrus et ai, 1975; Balaine et al., 1981; Tigges et al, 1984; Beaudry et al. 1988) to be positively correlated with profit in the US and Canada, the magnitude of the correlation depending on the measure of profit adopted. Examples of such estimates are; .47 for profit per day of life (Andrus et al, 1975); .17 for profit per day (Balaine et al, 1981), and .16 (Balaine et al, 1981) and .63 (Tigges et al., 1984) for net income per day of productive life. Beaudry et al. (1988) in a study of 176,902 Holstein cows found that lifetime total net income was highly and positively correlated with length of productive life and lifetime total milk yield. The correlation's were however influenced by the prices used to calculate feed costs and product values in the profit function. The association between length of productive life and profit is due to several factors: a) when cows have an increased length of productive life, replacement is delayed and hence replacement rate in the herd decreases and this decreases both replacement costs and cow depreciation costs (Allaire, 1981); b) cows increase production as they approach  59  maturity ( around third lactation), and the combination of longer length of productive life and decreased culling rate, creates an older herd on average; c) when involuntary culling is decreased, more voluntary culling is made possible, allowing potential (though small) improvement in milk production by improving the genetic material on the farm (Rendel and Robertson, 1950). The decision on whether or not to cull a cow at the end of a lactation is directly or indirectly influenced by production and thus profitability. Thus cows with low profitability, which could arisefrompoor performance in either production or fertility (for example multiple services per conception, abortion, still births and so on) will tend to have shorter productive lives. Jairath et al. (1995) reported phenotypic and genetic correlations estimates of .55 and .84 respectively between first lactation milk yield and length of productive life. The length of productive life is itself closely related to the number of lactations a cow is allowed to have and thus the relationship between number of lactations and first lactation yield would follow that with length of productive life. Jairath et al. (1995) found genetic and phenotypic correlation estimates of .79 and .46, respectively, between first lactation milk yield and the number of lactations an animal was allowed to have. The lifetime traits, length of productive life and number of lactations, however, had very low heritability estimates of .08 and .07, respectively. More than fifty percent of the cow's total energy intake in a production cycle is determined by its energy requirements for maintenance of body weight and growth (National Research Council, 1989). Consequently the largest fraction of variable costs of milk production is the one associated with feed costs and thus the relationship between cow size and net income or profit lies in its effect on rearing costs, cow maintenance costs (feed) and the salvage value. In most instances the difference between salvage value and the initial value at first freshening is negative (Pearson and Miller, 1981). Although larger cows tend to produce more milk (Ahlborn  60  and Dempfle, 1992; Hietanen and Ojala, 1995), it is unclear whether the advantage in milk production is enough to offset the difference in salvage and initial values. It has been reported that taller, longer and in particular heavier cows are less efficient in comparison to smaller cows (Sieber et al, 1988; Yerex et al, 1988). Dickerson (1970) calculated indexes to improve feed efficiency in the first lactation. He found that the index that included fat-corrected milk and weight change over the lactation was 33% more efficient than selection based on milk alone and also more efficient than selection based on feed efficiency alone. Simulation studies by Dempfle (1986) and Van Raden (1988) using profit functions showed a negative economic value for body weight, thus favouring reduced body size at a constant level of production. In contrast to the findings of Dempfle (1986) and Van Raden (1988), Gill and Allaire (1976) found a small but positive relationship between weight at first freshening and profit per day of life. Similar results were reported by Andrus and McGilliard (1975) who estimated the partial regression of profit on body weight at freshening to be $ .04/kg. This estimate was however not significant. In the study by Gilmore (1977) body weight was not considered, but body capacity score which could be regarded as a body size variable was studied. There was no significant relationship between body capacity and annualized net income. A significant negative relationship, however, was found between body capacity and annualized milk less feed and health costs. Body weight after calving was positively related to profit. Most dairy farmers do not attach a lot of value to type traits unless they have a direct effect on the milkability or health condition of a cow. Gilmore (1977) found that among the type traits he examined, only dairy capacity was significantly related to annualized net income. Rather than having a direct effect on net incomes, the effect of type traits appears to be realized through their influence on the herd life of cows (Vukasinovic et al., 1995; Pearson and Miller, 1988).  61  3.3.3. Other factors affecting profitability  Milk prices in most places are determined by the nature of the products the target market desires. Thus a market where all or most of the milk goes to fluid consumption, for example British Columbia, Canada, the prices are based on milk volume/weight with a milk fat differential. Milk markets where the larger proportion of the milk produced goes to processed products will pay based on milk fat with or without protein consideration. The New Zealand dairy industry sells a lot of its dairy products in the export market, particularly milk fat, thus the prices farmers receive are determined by the milk fat and protein content of the milk they deliver. These different values attached to the milk a cow produces could imply different levels of apparent economic efficiency. The effect of pricing systems on the outcome of economic selection indexes was addressed by Gibson (1989a, 1989b) who found that pricing systems have a profound effect on the responses, and the direction of response, to economic selection. This demonstrates the effect of price on the profit functions associated with milk yield and its components. Balaine et al. (1981) did a small study involving 182 Holstein-Friesian cattle at the Beltsville Agricultural Research Center. They used different prices to determine the effect of price increase on profitability ranking of cows. Little effect of change of price was found on the relative economic weights. These results were consistent given that the change in price was an increase rather than a change in the valuation system like that considered by Gibson (1989a). Similar results were found by Beaudry et al. (1988) who found little impact of prices on relative net income estimates. Reared in the same environment, different breeds might produce at different levels and achieve different liveweights. The question then arises whether the efficiencies and thus profitabilities of the breeds would also differ in a given value system. Visscher and Goddard  62  (1995) looked at Holstein-Friesians and Jerseys in the Australian dairy system and found that the average Income/Expense ratio for Jerseys was almost identical to that for Holstein-Friesians. The payment system used in the study, however, was based on yield of fat and protein with a penalty for volume. Stott and DeLorenzo (1988) compared the profitability of Holstein and Jersey cows in a herd in Florida. They found a significant effect of breed by season of calving interaction and that breed differences in some seasons of calving were significant. Holsteins were consistently superior to the Jerseys in profit. In a study of the effect of cow classification/grouping de Haan et al. (1992) compared grade and registered Holsteins for lifetime relative net income and relative net income adjusted for opportunity costs. Their results showed that the mean profitability measurements for registered cows were higher than those for grade cows. Similar results were found by Cassell et al. (1993) who found that in herds composed of grade and registered cows registered cows had significantly higher net income than grade cows and that the advantage increased with increase in herdlife opportunity.  3.3.4. Genetic parameters for profitability  Profit as a cow trait will vary in its nature depending on how it is defined. It could be perceived as a production trait, an efficiency trait or even an index whereby its derivation is based directly or indirectly on several different basic traits. Milk and milk component yields are the main traits in the computation of cow profitability. Some of the other traits included in the derivation will have major economic effects but have low heritability (for example reproduction, mastitis, and longevity) or moderate heritabilities but small net economic values (for example, growth rate, body size and milk flow rates), or are correlated strongly to milk yield traits (for  63  example, feed efficiency) (Pearson and Miller, 1988). The few heritability estimates for profitability available in literature are therefore as variable as the definitions of the profit functions used. Visscher and Goddard (1995) estimated heritabilities in Australian Holstein-Friesians and Jerseys for profit up to different stages of production and which were viewed as different profit traits. They defined their profit traits in the form of Income/food requirement ratio up to the beginning of a given lactation starting from the beginning of lactation two to the beginning of lactation six. They reported heritability estimates in the range .12 to .17 among the HolsteinFriesians and a range of .24 to .31 among the Jerseys. The estimates for lifetime profit were . 13 and .19 for Holstein-Friesians and Jerseys, respectively. Jairath et al. (1995) using a net income function reported heritability estimates .26 and .12 for profit up to the end of lactation one and lifetime profit, respectively. They also found heritability estimates of .29 and .03 for profit per day of first lactation and profit per day of productive life respectively.  3.4. MATERIALS AND METHODS  3.4.1. Data  The data used was from the CANZ trial referred to in Chapter 1. The data collected for this part of the study included pedigree information, dates of birth, calving and disposal, lactation yields of milk, milk fat and protein in kilograms. Also collected was information on liveweights and reasons for disposal. The original data consisted of 834 lactation records from 343 cows in Canada obtained from the Canadian Dairy Network and 3621 lactation records from 834 cows in New Zealand obtained from the New Zealand Dairy Board. In Canada all cows that  64  did not have any weight records after the first calving and those whose growth function failed to converge as explained in Chapter 2 were edited out. Also edited out were those cows with missing production information for example milk, fat or protein yield, number of days in milk and calving dates. Any cow without a completed first lactation was edited out. In this respect any first lactation with less than 100 days milked was considered incomplete. The same editing criteria were used in New Zealand. Four herds in Canada and one in New Zealand had incomplete liveweights data for all the project cows they had and these were all removed from the analysis. After all the edits the data remaining consisted of 596 lactation records from 250 cows in 6 herds in Canada and 2811 lactation records from 670 cows in 19 herds in New Zealand. In New Zealand all the cows were born in the 1985 calving season while in Canada cows were bom over the period 1985 to 1989 and calving was all the year round.  3.4.2. Definition of the economic efficiency trait  As explained in Chapter 1, the project cows in either country were reared under the normal management practices of each herd. The data was collected under field conditions and no actual feed consumption information was available. No information was recorded either on health matters. The economic efficiency trait used in this study was of the form of an Income/Expense ratio, one of the measures of profit described in Section 2.2.1. However rather than expense being expressed in monetary value as income was, it was expressed in terms of food requirements in the form of Metabolizable Energy (ME) requirements. The reasons this form of profit measure was adopted in preference over any of the other measures described in the literature reviewed in sub-section 3.3.1 are;  65  1) The data in this study was collected under field conditions in both countries and no feed intake data was available. The use of the ratio Income/Energy Requirements as the measure of profit provides a solution to the problem of putting a cost on feed consumed since actual consumption information was not available.  2) In New Zealand the cows were totally dependent on pasture for their nutritional requirements. In dairy systems where feeding is based on pasture, the total cost of feed is dependent on the size of farm, quality of land and the level (therefore cost) of pasture management among others. To put a cost on feed one would have to give a cost to all the factors mentioned here, but the problem becomes what cost/price to attach to land when even the areas under pasture in the farms were not available? A similar question also arises for Canada since most dairy farms produce their own forage on the farm.  3) Brascamp et al (1985) demonstrated that relative economic weights are the same, whether profit is expressed per cow or per unit of food (or energy). The components of the ratio, returns and energy requirements, were computed for each cow up to the end of each completed lactation. Let $Rj and FRj be the returns from milk produced and food requirements in terms of Metabolizable Energy (ME) up to the end of lactation i where:  FRj=£  (MEmain + MEgain + MEpreg + MEmilk)  (3.1)  I  where; ME ain is the Metabolizable Energy requirement for maintenance; M E g m  arn  is the  66  Metabolizable Energy requirement for weight gain; M E p  reg  is the Metabolizable Energy  requirements for pregnancy and MEmilk is the Metabolizable Energy requirement for the milk produced over lactation i, and  $R; =  i  M x Price,-  (3.2)  where; M is the cumulative milk yield from day of first freshening, or a function of milk, fat and protein; Pricej is the price paid per unit of milk yield or the function of components thereof in the jth milk market;  3.4.2.1. Estimating food (Metabolizable Energy) requirements  As stated earlier the food requirements FR, were computed in the form of metabolizable energy, ME. This was necessitated by the fact that no feed consumption records were available for either of the countries. While it would be possible to estimate feed costs for the Canadian environment by the assumption of a given diet composition and prices of feedstuffs as done by Harris (1992), this would be difficult to do for New Zealand where cows were grazed on pasture. Pearson and Miller (1981) suggested that where feed consumed has not been recorded NRC specifications could be used to compute a profit or income function. In this study the energy requirements were worked out per lactation based on a cows milk production and liveweight over that lactation. The energy equivalent of milk or the net energy it contains is mainly determined by the concentrations of fat and protein. Dado et al. (1990) derived theoretical equations for calculating the energy requirements for milk production. The equations estimate the minimum requirements of metabolizable energy for production of milk fat, protein and  67  lactose. The equations do take into account the Adenosine Triphosphate (ATP) and amino acid that are used between the absorption of substrates and final products. With the assumption that 5% of the glucose requirement is provided by amino acids Dado et al. (1990) gave the metabolizable energy requirements in Meal as;  MEmilk = 13.42F + 5.98L + 7.60P  (3.3)  where F is the fat yield, P the protein yield and L the lactose yield for the lactation. The amount of lactose in milk is fairly constant due to the contribution of lactose to the osmotic value of milk (Dommerholt and Wilmink, 1986), and since no data on lactose production was available, the lactose concentration in this study was assumed to be 4.96% (Harris 1992. Equation (2.1) can therefore be written as;  MEmilk = 13.42F + 0.297M + 7.60P  (3.4)  The daily metabolizable energy requirements in Meal for pregnancy were computed as;  MEpreg = 0.04LWT 0  75  (3.5)  where LWT is the liveweight of the cow in kilograms (NRC, 1989). This was computed for the last 73 days of gestation. The metabolizable energy requirements for maintenance are expressed in terms of the 0 75 metabolic liveweight given as LWT ' . The equations for computing metabolizable energy requirements in Meal by parity for lactating cows are (NRC, 1989);  68  parity 1  ME  m a m  = 0.1600LWT -  75  (3.6)  parity 2  ME  m a m  = 0.1467LWT '  0 75  (3.7)  parity 3+  M E i = 0.1333LWT  0  m a  (3.8)  0,75  n  For dry cows, which are assumed less efficient in the energy utilization than lactating cows, the equations are (NRC, 1989);  parity 1  M E i = 0.2043LWT  0-75  (3.9)  parity 2  ME fn = 0.1873LWT  0,75  (3.10)  parity 3+  M E i = 0.1702LWT "  m a  n  ma  0  m a  (3.11)  75  n  The energy requirements for parities 1 and 2 need also to include the energy required for growth since the cows are still not mature at that time. The energy required for the gain in weight was computed from NRC (1989) equations which assume an efficiency of metabolizable energy of 0.60 and 0.47 for lactating and dry cows respectively. The energy was computed as;  lactating  MEgain = 0.0583LWT°- LWG  dry  M E g i = 0.0744LWT°- LWG  75  75  a  n  1-119  L119  + 1.7LWG  (3.12)  + 1.7LWG  (3.13)  where LWG is the liveweight gain per day in kilograms.  69  3.4.2.2. Liveweights  The liveweights of the cows in the study were taken regularly particularly at the early ages. In New Zealand the last weighings recorded were at age 30 months approximately six months after first calving. In Canada many of the herds continued weighing the animals although not on at regularly or evenly spaced periods. The liveweights at calving were needed in this study to estimate the energy requirements for each cow in a given lactation. Since these weights were not always available due to the nature of the way the data was collected a growth function was needed to estimate and plot the growth curve of each cow, and thus obtain its estimated liveweight at any given point. In Chapter 2 it was explained how four different growth function were fit to the weight data of individual cows involved in the study. Of the four growth functions tested the Von Bertalanffy was chosen for this part of the study since it was found to give the best overall fit to the weight data of the cows in both environments. This function has the form;  (3.14)  where; LWT is the live weight at age t (days), A is the asymptotic mature weight, b is a constant t  of integration which indicates the proportion of the asymptotic mature weight to be gained after birth, and k is a maturation index The growth curves obtained after fitting this function were then used to estimate liveweights at each calving for the cows in the study. As explained in Chapter 2, cows were edited out if their weight data failed to converge using the function at the specified 50 iterates. In both Canada and New Zealand cows with no weight records after the first calving were eliminated from the analysis. After these and other criteria for editing out as explained in section  70  2.3.1, the number of cows remaining for further analysis were 250 in Canada and 670 in New Zealand. Since the dates of birth and of each calving were known for each cow the weights at first and second calvings were estimated using each cows parameter estimates in the function as given above. The weights at third and later calvings were assumed to be approximately equal to the asymptotic mature weight A.  3.4.2.3. Returns  In Canada returns were computed based on two systems of milk pricing. The first is the pricing system based on fluid and milk fat as used in British Columbia. The price used in this study was the weighted average price paid for milk within and above quota in the accounting year 1995/96 (B. C. Milk Marketing Board, 1996). The milk value was computed as;  Price = 35.81(SM/100) + 0.4585(F/10) BC  (3.15)  where SM is the lactation yield of skim milk (whole milk less milk fat) in liters and F is the lactation yield of milk fat in kilograms. The value is in Canadian dollars. The second pricing system used to get the returns for milk is the component system used in the province of Ontario where fat, protein and other solids are considered. The value of milk was computed based on the average price paid in the 1995/96 dairy year (Attaher Maiga; personal communication) as;  Price = 5.4798F + 7.6241P + 0.6587OS 0N  (3.16)  where F is the lactation yield of milk fat, P is the lactation yield of protein and OS the yield of  71  other solids in kilograms. Since the records acquired for the study did not provide the yields of other solids, it was assumed in this study that the other solids component was primarily composed of lactose, which was further assumed to be 4.96% of the total milk yield. In New Zealand milk returns were computed based on the price paid by the New Zealand Dairy Board (Vanessa Smith, personal communication) in the 1995/96 season. The value in New Zealand dollars was computed based on lactation yields of milk fat and protein as;  Price = 2.61F + 5.08P  (3.17)  NZ  where F and P are the lactation yields of fat and protein, respectively. After the computation of the returns and requirements for energy, the ratio of the two was used to compute the profit trait which was defined as;  PFij = SRij/FRjj  (3.18)  where PFjj is the measure of economic efficiency for the jth cow up to the end of the ith lactation. The economic efficiency trait PFjj was computed for each cow in the dataset up to the end of its first lactation and up to the end of its last lactation, that is, over its entire productive life from day of first freshening to the last day of the last lactation, regardless of its fate. Thus for each cow two estimates of economic efficiency were obtained. 1) PF,j = efficiency up to the end of lactation 1, 2) PF = efficiency over entire productive life , Lj  Hence for a cow that had only one completed lactation PF, and PF were equivalent. L  72  3.4.3. Analyses  As stated earlier economic efficiency for each individual cow was computed for two production spans, over the first lactation (PFi) and the over the entire productive life (PF ). The L  two profit traits and first lactation yield traits of milk, fat and protein, and the percentage yields of fat and protein were analyzed to test for the effects of the fixed factors of strain of sire and herd and the random effect of sire within strain. The mixed model analysis was done using the GLM procedure of SAS (SAS Institute Inc., 1990). The analysis was done within environment using the model,  Yijklm = H +Ti + Sj(Ti) + H + e k  ijk  (3.19)  where; Yyk  = an observation on the ijk cow,  Tj  = the fixed effect of the ith strain of sire (Canadian or New Zealand)  tn  Sj(Tj) = the random effect of the jth sire within strain i, H e  k  ijk  = the fixed effect of the k ^ herd-year-season (Canada)/herd (New Zealand), ~^  e ran  dom error associated with the ijkth observation.  For the first lactation yield traits the number of days in milk was included in the model as a covariate to adjust for number of days in milk. The heritabilities of the profit traits were estimated using Derivative-Free Restricted Maximum Likelihood (DFREML) (Meyer, 1991) with an animal model. The model used was;  y=Xb + Za + e  (3.20)  73  where y is a vector of observations (cow economic efficiency, milk yield, and so on), b is a vector of fixed herd and strain effects, a ~ (0, Aa ) is a vector of additive genetic values, e ~ (0, 2  a  Ia ) is a vector of errors and X and Z are known design matrices of the fixed and random animal 2  e  effects respectively. A is the additive genetic relationship matrix which includes cows, sires, dams and maternal grandsires and I is the identity matrix. Correlations between the random effects were assumed to be zero. Additive genetic and phenotypic variances were also obtained from the same model. Genetic and phenotypic correlations between the economic efficiency traits and first lactation yield traits of milk, fat and protein were estimated for the New Zealand environment using the bivariate equivalent design animal model of DFREML (Meyer, 1991).  The  correlations in the Canadian environment could not be reliably estimated due to convergence problems with the bivariate model because of the very sparse observations in the dataset. Standard errors for the genetic correlations were calculated using the method given by Falconer (1989). The relationships between first lactation production traits and lifetime economic efficiency will assist in determining the most efficient of the first lactation traits to use in the prediction of sire lifetime profitabilities. The heritabilities, and genetic correlations between lifetime profit and first lactation traits, were used to calculate predicted response to direct selection for lifetime economic efficiency and correlated responses to indirect selection through selection for first lactation traits of milk, fat and protein yields and for economic efficiency(PF,).  74  3.5. RESULTS AND DISCUSSION 3.5.1. First lactation yield traits  The analysis of variance results (R x 100) for first lactation production and economic 2  efficiency traits for the Canadian environment are presented in Tables 3.1. In Canada the effects of sire within strain and herd-year season of first calving were important for all lactation one yield traits. A significant effect of sire within strain indicates that even though the sires selected were among the best in their respective country, there still are considerable differences among them in the Canadian environment for all of the traits tested. Strain of sire had a significant effect on lactation milk yield, and percentage of fat. The least squares means for two strains in Canadian herds are presented in Table 3.2. Daughters of Canadian bulls produced a mean of 7020 kg of milk with a mean fat percentage of 3.63 compared to a mean of 6845 kg with mean fat percentage of 3.77 by daughters of New Zealand bulls. These results agree with those of Peterson (1991) who reported significant differences between the daughters of the two strains of sires in standardized 305-day first lactation milk yield and in fat and protein percentage.  75  * ft  hfl  *T)  ft "rt  ~  33  <  en  a.  y  o CL GO  3  o  I  rtP o  P  3  . 3' 1=1  3.  3  o  o » o  3  35  HH  EL Ul  Cu  00  i » 2, ^  » "3 p ^-  CD  cr:  3  org  3  »  o S. sr O 3 O 3 Cu 3  s.  < o TJo  §• 3. rt ^3 3^ '-i  3. o a  ON  4*.  r  a  Ul *  ft  P  CO  o a  ~J  bo *  — ik NO  #  i—» NO  *  *  o  4^  to  L)  NO ON  to  "3  o I-I  rt  3'  £1 to  ON  o  i—»  ON U>  3 °  rt  Ul  O  3?  o  o +  ~J  to  o ©  © to  o 3  o z  +  s,  3  p  3  o ft o ft'  T)  o  v; PT P  r 1  CL  r o  o p  3 OQ o  ui Ul  -0 NO  NO  bo  Ul to  to NO *  >  p  rr ft ON ON  I-  16.8  1  to  O  3  9  "T3  10.2  >-t rt 2 3  28.2  00  a ^  O  to -o ON  3 °  "  o  8  a ^  s  ft o  O  §  o rt  p  CL  o  35 g 2.  16.4  0  oo  10.8  O  27.5  o 3  p  ©  +  NO  v  o  o o  P  o  3  X  1' u>  y Ti  U)  27.6*  I  II  cr. o  13.2  ^  o ft  3' 23.8*  S  ©  o  53.2*  t-  1  I'  0.0+  CL  19.9*  P  <  C  9.6*  c-t-  rt  o 3  14.3  o  83.0*  o  21.6*  si 3* 2o P ft  13.6*  ft  33  80.2*  3  ^  16.3  O O  T3  0.4*  ss  16.9*  o g P v: o  2  o 9.5*  A  80.9*  ft'  15.2  ST °  to Ul  O oo bo  E3  P CL  3"  76  Table 3.2. Least squares means and standard errors by strain of sire for first lactation and lifetime traits in the Canadian environment. Canadian (n = 137) New Zealand (n = 113) Mean (s.e.)  Mean (s.e.)  Milk yield  7020* (135)  6845 (154)  Fat yield  252 (5)  254 (6)  Protein yield  229 (4)  229 (6)  Fat %  3.63 (.06)  3.77 (.07)  Protein %  3.28 (.03)  3.32 (.03)  Calving interval  425 (19)  377 (24)  FR,  14.3 (0.4)  14.5 (0.4)  $R l(ON)  3388 (101)  3482 (120)  $R 1(BC)  3798 (113)  3829 (133)  PF l(ON)  235 (2.3)  235 (2.7)  PFl(BC)  261 (2.8)  259 (3.3)  38.2 (2.5)*  29.5 (3.0)  $R-L(ON)  8996 (614)*  6978 (725)  $R-L(BC)  10403 (727)*  8022 (845)  PFL(ON)  236 (2.3)  235 (2.7)  PFL(BC)  267 (2.9)  262 (3.4)  NLACT  2.5* (0.2)  2.2 (3.4)  LPL  864* (69)  703 (80)  FR  L  * differences between strains are significant (P < 0.05) PF , = economic efficiency up to the end of lactation 1 with Ontario prices in $/l,000 Meal, PF , = economic efficiency up to the end of lactation 1 with BC prices in $/1,000 Meal, PFHON) economic efficiency over productive life with Ontario prices in $/l,000 Meal, PF = economic efficiency over productive life with BC prices in $/1,000 Meal, NLACT = number of completed lactations, LPL = length of productive life, FR, = metabolizable energy requirements over first lactation in 000,s of Meal, $R,(ON) returns overfirstlactation in Can$ using Ontario prices, $RI( Q returns over first lactation in Can$ using BC prices, FR = metabolizable energy requirements over productive life in 000,s of Meal, $R = returns over productive life in Can$ using Ontario prices, $R = returns over productive life in Can$ using BC prices, 1(0N)  1(BC)  =  L(BC)  =  =  B  L  L(0N)  L(BC)  Analysis of variance results (R x 100) for the New Zealand environment are presented 2  in Table 3.3 and the least squares means in Table 3.4. In New Zealand the effects of sire within strain and of herd were important sources of variation for all lactation one yield traits again indicating important differences among the sires themselves irrespective of the selection. Strain of sire was significant for first lactation milk and protein yields and percentages of fat and protein. The overall mean first lactation yields of milk, fat and protein were much lower than in Canada. The overall mean yields were , 6933, 253 and 229 kilograms in Canada and 3247, 137 and 106 kilograms in New Zealand for milk, fat and protein, respectively. Daughters of Canadian sires produced significantly more milk in the first lactation than their New Zealand sired counterparts with means of 3376 and 3117 respectively. Consistent with the results in Canada, no significant difference was found between first lactation fat yield. There was, however, a significant difference in protein yield with the daughters of New Zealand bulls producing approximately 4 kilograms more. The mean percent fat and percent protein among daughters of New Zealand sires were significantly higher than among daughters of Canadian sires while in Canada a significant superiority was only realized in fat percentage. In general and as expected due to the smaller dataset the standard errors were bigger in the Canadian data. The significant superiority of daughters of Canadian bulls in both environments for first lactation milk yield was different from the findings of Bar-Anan et al. (1987) in the Israeli experiment.  They found no significant difference between Canadian sired cows and New  Zealand sired cows in mean first lactation milk yield. In both environments in this study, and similar to the findings in the study of Bar- Anan et al. (1987), the percentage of fat in first lactation milk was significantly higher among New Zealand sired cows than Canadian sired ones.  78  £rt  f3  a>  o  CL  (!) 2  ll  3  o  K CO  l-t Cu  B P 3'  «'  2.  I  oo CD  B p 3'  ft o  O J  o  O J  EL  §- « ai ft ^  ft a  o P  O  Bl  3  O 3  a "§ Cu O  1—I  00  OJ  00  s.  rt  Ul  oo  pi Ui  Ha"  Bi O ft p ft B. 3 o a-  » s, t- p* 1  to  4* #  Os i—'  bo  ©  *  to p Ul  *  to  4^ bs  *  to bo  B  o  o ft  *  1-1  ft  CO I—»  P  4^  N)  *  *  II B .  5T B 3 o  1—  < s. ri o  OJ  SO  Ul  *  r  II o o a- o c 3 o O  Os  to to  so OJ  *  to *  *  & i-t Bl  3'  p o  to  so  O O  TJ o ft  Os *  to  S"  Ul  3  ft  8 o o <;  Os Ul  OJ  4^ 4^ #  OJ  SO  *  i—'  to SO  *  o Ul  *  3'  to Ul  4^ #  so 4^. *  o  Ul  ©  Ul  Ul  *  OJ  o  so *  3  B P  ft Ul  O  o fc i-h Bl o a' 3 o  a o a. c  p  Cu ft) o o 3  SO bo  *  © *  TJ Tj  p  Cu o  £L <f OJ OJ  4^  4^  00  to  SO  to  *  *  © *  TJ Tj  3'  OQ  3' rf  a 3 Os O 4^  Os SO  ©  OJ  © *  to SO  *  4^  Os OJ  4^ 4^  Os 4*  to Ul  #  to Ul  *  > O  H TJ  r  1  ft ft  N ft p  I* Cu  7 9  Table 3.4. Least squares means and standard errors for first lactation and lifetime traits in the New Zealand environment. Canadian (n = 330) New Zealand (n = 340) Mean (s.e.)  Mean (s.e.)  Milk yield  3376* (30)  3117(33)  Fat yield  138(1)  136(1)  Protein yield  108* (1)  104(1)  Fat %  4.08* (.02)  4.37 (.03)  Protein %  3.20* (.01)  3.32 (.01  376 (2)  372 (3)  FR,  8.0 (0.1)*  7.8 (0.1)  $Ri  957 (10)*  931 (11)  PF,  119(0.5)  120 (0.6)  FR  L  29.9 (17)*  34.9(1.8)  $R  L  3595 (201)*  4294(218)  PFL  121* (0.6)  123 (0.6)  NLACT  3.8* (0.2)  4.5 (0.2)  LPL  1214* (61)  1456 (66)  Calving interval  * differences between strains are significant (P < 0.05) PF, = econmic efficiency up to the end of lactation 1 in $/1,000 Meal, PF = econmic efficiency over productive life in $/1,000 Meal, NLACT = number of completed lactations, LPL = length of productive life. FR! = metabolizable energy requirements overfirstlactation in 000,s of Meal, $R = returns over first lactation in NZ$ FR = metabolizable energy requirements over productive life in 000,s of Meal, $R = returns over productive life in NZ$ L  1(0N)  L  L  No differences were found between sire strains for first calving interval (calving one to two) in either country. Cows sired by Canadian sires, however, had slightly longer calving intervals in both environments. They had a calving interval 42 days longer in Canada and 4 days longer in New Zealand. In New Zealand calving interval showed much less variability, possibly due to the management practice of seasonal breeding, as the relatively low standard errors show. It is therefore likely more influenced by management practices in that environment than it would be in Canada. Management not withstanding, the effect of the cow itself would however still be important due to its influence on fertility, and specifically the success or failure of conception (days open). The tendency in New Zealand is that if a cow does not conceive within the designated mating season, and she is a good cow, the farmer may leave the cow open until the following breeding season. Otherwise she is culled at the end of the lactation. It was therefore possible to get cows with extremely long calving intervals, but such cows (42) were excluded from the analyses.  3.5.2. Economic efficiency and lifetime traits  In Canada the Ontario and British Columbia (BC) price systems were considered in computing economic efficiency. No significant effect for strain of sire were found under either pricing system for any of the two economic efficiency traits in Canada. This lack of a difference between strain of sire in economic efficiency traits in Canada mirrors the findings of Graham et al. (1991) who worked with 65 of the cows involved in this study from one herd in Ontario (University of Guelph). They found that differences in milk returns between strains when computed using Canadian (Ontario) and New Zealand payment systems were not significant. They also found no significant difference between sire strain in energetic efficiency, which was  81  defined as 305-day lactation energy output/feed energy intake. Neither sire within strain nor herd-year-season were significant for any of the economic efficiency traits. It should be noted that even though there were no differences in the economic efficiency traits between daughters of the two strains in Canada, there were significant differences in the energy requirements and in the returns over productive life. The differences in returns were realized for both Ontario and BC milk pricing systems. In both systems daughters of Canadian sires had higher returns. This higher returns were, however, accompanied by higher energy requirements and the resulting ratio (economic efficiency) was not significantly different between the two strains. In New Zealand strain of sire was important for economic efficiency over the productive life (PF ) but not for economic efficiency up to the end of lactation one. Daughters of New L  Zealand sires were more efficient (123 $/l,000 Meal) over their productive lives than the daughters of Canadian sires (121 $/l,000 Meal). Both sire within strain and herd were significant effects for both economic efficiency up to the end of first lactation and lifetime economic efficiency. Consistent with the observations in Canada, there were significant differences in the energy requirements and in the returns between the two strains in New Zealand. Contrary to the results in Canada, however, daughters of New Zealand sires had higher requirements and higher returns over their productive life than daughters of Canadian bulls. The same trend was observed over the first lactation. The differences over productive life, like in Canada, could be an artifact of the data due to the fact that in each country daughters of native bulls had longer lengths of productive life (LPL) and more number of lactations completed (NLACT). The ratio is therefore still the more fitting criterion of comparison since it gives a measure of the estimate of each cow's efficiency regardless of the length of productive life.  82  What is worthy of note in this study though, are the differences between the strains in the economic efficiency traits with the different pricing systems. In Canada, using Ontario prices, the daughters of both strains of sires were similar in both economic efficiency traits. This was also the case in New Zealand for economic efficiency over the first lactation but not for economic efficiency over productive life. Both these systems value milk based on both fat and protein yields. Fat yields were not significantly different in either environment, but protein was in New Zealand. With BC milk prices, Canadian sired cows had the slight advantage, though insignificant.  As seen earlier the BC pricing structure pays for volume with a milk-fat  differential but does not pay for protein since this is considered primarily a fluid market. Since Canadian sires have mostly been selected for milk yield this could have resulted in the observed change of ranking of the sire strains with respect to economic efficiency. In Canada the denominator for the economic efficiency function ($R/ FR) was the same across both payment systems, implying the difference observed in economic efficiency was actually due to differences in revenue, and thus the value systems for milk between the two provinces. The group of project sires from New Zealand were the top sires for fat yield, and although their daughters were significantly lower in milk yield they compensated with higher percentages of both fat and protein, and in New Zealand a higher protein yield.  This might explain the significant  differences seen in economic efficiency of productive life with the New Zealand pricing system that gives more weight to fat and protein. This significant difference in lifetime economic efficiency could also be partly due to the fact that New Zealand cows had on average more completed lactations. Mean cow economic efficiency appears to increase with increasing lactations, at least up to lactation three. In general cows in New Zealand had more lactations completed (NLACT) and a longer mean length of productive life (LPL) than cows in Canada, implying a higher herd turnover rate  83  in Canada. The overall means for NLACT and LPL were 2.4 lactations and 784 days in Canada and 4.2 and 1335 days in New Zealand. In both environments there was a significant effect of sire strain on NLACT and LPL. In Canada, cowsfromCanadian sires had more lactations (2.5 vs. 2.2) and had a LPL 161 days longer (864 vs. 703). The reverse was observed in New Zealand where cowsfromnative bulls had on average more lactations (4.5 vs. 3.8) and 242 days longer in LPL (1456 vs. 1214). It is not clearfromthe production results why this would be the case. Indeed in New Zealand, Canadian sired cows had on average significantly higher first lactation yields of milk and protein. In Canada, Canadian sired cows also produced significantly more milk. In both environments economic efficiency was also quite comparable. If the differences in productive lifespans were influenced by health or fertility reasons, it is not clear from the information available. Disposal data showed little differences infrequenciesof culling for the various reasons between daughters of the two strains (Mwansa, 1997). No reason is therefore discernible that can adequately explain the overall differences in NLACT and LPL.  3.5.3. Heritability estimates  Estimates of heritability for first lactation and economic efficiency traits in both Canadian and New Zealand herds are presented in Table 3.5 while the phenotypic variances are presented in Table 3.6. Estimates for first lactation yields of milk, fat and protein were close to those reported by Jairath et al. (1995), but higher than those reported by Gill and Allaire (1976) in similar studies looking at the relationship of first lactation production traits and lifetime performance. The estimates in Canada were however lower than the estimates reported by Peterson (personal communication) for a subset of the dataset. These lower heritabilities could  84  Table 3.5. Heritability estimates (and standard errors) for first lactation and economic efficiency traits in Canada and New Zealand Environment Canada  New Zealand  Milk yield  .28 (.18)  .26 (.11)  Fat yield  .28 (.18)  .14 (.09)  Protein yield  .17 (.18)  .24 (.10)  Fat %  .54 (.35)  .26 (.11)  Protein %  .33 (.21)  .24 (.11)  P F l (ON)  .12 (.17)  PFI(BC)  .15 (.17)  PF,  .20 (.110)  PFL(ON)  .21 (.20)  PFL(BC)  .15 (.18)  PF  L  .34 (.14)  85  Table 3.6. Phenotypic variances for first lactation and economic efficiency traits in Canada and New Zealand Environment Canada  New Zealand  5736451  404364  Fat yield  8475  703  Protein yield  6070  423  Fat%  0.361  0.164  Protein %  0.047  0.051  Milk yield  PF, ON)  448  (  PF  652  1(BC)  PF,  —  PF N)  428  L(0  PF PF  86  690  L(BC)  90  —  L  FR,  12.2  1.75  FR  512.5  681  L  $R, ON)  9  7  9  2  5  6  (  $R  1202786  1(BC)  $R,  —  $R  L(0N)  30562379  $R  L(BC)  43126842  $R  —-  L  39472  10664762  FR, = Metabolizable energy requirements for first lactation, FR, = Metabolizable energy requirements over productive life, R, = returns over first lactation, R = returns over productive life L  be explained by the fact that in the current study, lactation yields were used without standardization of lactation lengths as was the case in the study by Peterson, who was also able to use more of the records than could be used in this study. There was considerable variation in the lactation lengths in the data set most of which would be non-genetic, thus lowering the estimates of heritability. The standard errors relative to the estimates were much higher in Canada due to the lower number of records in the analysis. Heritability estimates for percent fat and percent protein were .26 ± .15 and .24 + .13 in New Zealand and .54 + .35 and .33 ± .21 in Canada respectively. These estimates are within range of the estimates given in literature. Heritabilities of economic efficiency traits were higher in the New Zealand than in the Canadian environment. In New Zealand heritability for economic efficiency over first lactation was .20 ±.11 while in Canada the estimates were .12 ± .17 and .15 + .17 with Ontario and B. C. prices, respectively. The estimate in New Zealand is similar to that reported for profit by Jairath et al. (1995) (.26). In Canada the estimate is similar to the heritability estimate of .12 reported by Visscher and Goddard (1995) using a similar measure for Australian Holsteins but slightly higher than the 0.1 approximated by Klassen et al. (1992) for milk value/cost of food among Canadian Holsteins. Visscher and Goddard (1995) computed their profitability (returns/energy requirements) up to the beginning of lactation two, therefore including the dry period. In this study however, economic efficiency over first lactation was computed to the end of the lactation, thus excluding the dry period. The trait they worked with could therefore be interpreted as being slightly different. Lifetime economic efficiency, however, does include the dry periods in between subsequent lactations. However, generally the estimates in both environments in this study are lower than other estimates reported elsewhere (Gill and Allaire, 1976; Lin and Allaire, 1978). However, the estimates in the literature for first lactation profit vary quite a lot. This could be due to both differences in populations studied and differences in the computations of  87  the profit function. The profit functions used in the latter two studies of Gill and Allaire (1976) and Lin and Allaire (1978) were much more comprehensive as they included many of the cost and income variables explained in subsection 3.2.1. In New Zealand the estimate for lifetime profit was slightly higher than that for profitability to the end of lactation one at .32 ± .13 a pattern that was also observed in Canada using Ontario prices. This pattern differs from that of Jairath et al. (1995) who found a heritability estimate for lifetime profit (net income) much lower than the estimate for profit up to the end of the first lactation (.12 vs. .22). The pattern of estimates observed here was, however, close to that found by Visscher and Goddard (1995) who, working on Australian Friesian and Jerseys found similar estimates for heritability for profit (Income/Food Expense ratio) over a lifetime and profits up to the beginning of lactations 2,3,4, 5 and 6 for both strains. This would imply that the traits are closely related due to part-whole relationships. A similar reasoning would apply to the observed estimates in the current study, but no reason is clear why the heritability for profit over productive life is slightly higher that over lactation one.  3.5.4. Genetic and phenotypic correlations (New Zealand)  Genetic and phenotypic correlations were only estimated in the New Zealand environment. The small number of records in Canadian herds precluded their estimations for the Canadian environment.  The genetic and phenotypic correlations between first lactation  production traits and the two economic efficiency traits are presented in Table 3.6. The standard errors for the genetic correlations were relatively large, an outcome largely attributable to the small number of records that were available. Genetic correlations between the first lactation yield traits of milk, milk fat and protein and economic efficiency over the first lactation were  88  Table 3.7. Estimates of genetic correlations (above diagonal) and phenotypic correlations (below diagonal), between first lactation production traits and economic efficiency traits in the New Zealand environment (standard errors of genetic correlations given in brackets). Milk Fa! Protein PF\ PF\ Milk  .90 (.07)  .94 (.03)  .65 (.19)  .44 (.24)  .88 (.08)  .70 (.20)  .70 (.19)  .70 (.17)  .57 (.20)  Fat  .89  Protein  .94  .94  PF,  .73  .76  .78  PF  .54  .56  .58  L  .73 (.16) .80  all standard errors for phenotypic correlations were < .01  moderately high with estimates of .65, .70 and .70 for milk, fat and protein yield, respectively. Generally the estimates in this study were lower than those reported by Jairath et al. (1995) and Visscher and Goddard (1995).  The phenotypic correlations between the yield traits and  economic efficiency for the first lactation were moderately high at .73 to .78 Genetic correlations between the three yield traits and lifetime economic efficiency (PF ) were positive and in the medium range, from .44 for milk yield to .70 for fat yield. Thus L  first lactation fat yield had the highest correlation with lifetime profit of the three traits. This is perhaps not surprising given that the payment for milk is based on the yields of fat and protein in New Zealand. The correlation with protein was however lower at .57. The corresponding phenotypic correlations for the three first lactation yield traits and lifetime economic efficiency were all similar, .54 to .58. The phenotypic correlations for first lactation yield traits and lifetime economic efficiency in this study are closer to those reported by Visscher and Goddard (1995) (.57 to .60) in a similar study with Australian dairy cattle using a similar mesaure of  89  efficiency (profit) , but much lower than those reported by Jairath et al. (1995) (.82 to .88) in a study of Canadian Holsteins. The genetic correlations are also lower than those reported in the two studies. Lin and Allaire (1978) reported similar correlations between first lactation milk yield and lifetime profit. Both genetic and phenotypic correlations between economic efficiency over first lactation (PF,) and lifetime economic efficiency(PF ) were high (.72 and. 80 respectively). Both L  correlations were higher than those seen with first lactation yield traits, implying that selection for first lactation economic efficiency would result in a greater response in lifetime economic efficiency than selection for yield traits. The genetic correlation between first lactation profit and lifetime profit in this study is slightly lower than that reported by Jairath et al. (1995) who reported an estimate of .85, while their phenotypic correlation was much lower at .55 as compared to the .80 in this study. Direct response to selection for lifetime profit (PF ) assuming a selection intensity of 1.0 L  was estimated at 3.2 $/l,000 Meal. Estimates for correlated responses to selection for first lactation yields of milk, fat and protein were approximately 1.2, 1.4 and 1.5 $/l,000 Meal, respectively. The correlated responses to selection for PF! was approximately 1.8 $/l,000 Meal. Selection for lifetime economic efficiency would have the drawback of long generation intervals since breeders would have to await the culling of daughters, regardless of the length of their lives.  The use of first lactation traits would reduce this and thus increase the pace of  improvement. Among first lactation yield traits protein yield gave the highest correlated response but there was little difference in selecting for any of the three yield traits. The best correlated response would be realized with the selection for economic efficiency up to the end of lactation one, though it was only marginally better. This is strengthened by the high genetic correlation between the two traits (.73).  90  3.6.  CONCLUSIONS  Daughters of Canadian bulls yielded significantly more milk in the first lactation than daughters of New Zealand bulls in the Canadian environment.  There was, however, no  significant difference between them in the yields of fat and protein. New Zealand sired cows had a significantly higher percentage of protein in first lactation milk but fat percentage was not different between the daughters of the two sire strains. In New Zealand daughters of Canadian sires produced significantly more milk and protein than daughters of native bulls. Fat yield was, however, not significantly different. These differences translated into significantly higher mean percentages of fat and protein among daughters of New Zealand bulls. The lack of large differences in production and weights (as seen in Chapter 2) generally resulted in similar economic efficiency (returns/food requirements) between daughters of the two strains of bulls in both environments. However in New Zealand daughters of native bulls had higher economic efficiency over their productive life than the daughters of Canadian bulls. This could partly be attributed to the fact that the New Zealand sired cows had on average more completed lactations (4.5 vs. 3.8) than daughters of Canadian sires. A profitable lifetime is the goal farmers aim for and in New Zealand it appears that overall the more profitable way is to use local bulls. The importation of more expensive germplasm (Canadian) in the form of semen in this case would only be justified if the bull of choice has proofs (in New Zealand) that are higher than the local bulls. This is supported by the fact that the effect of sire within strain was significant for all economic efficiency traits. The profitability indicator used in this study however, was not exhaustive due to limitations on data collected. Many variables for example heifer rearing costs, health costs, semen costs, salvage  91  value and so on were not included and it is not clear whether this would change the nature of the results. Canadian sired cows in Canada had on average a longer length of productive life and were allowed to complete more lactations than their New Zealand sired contemporaries. The scenario was different in New Zealand where daughters of native bulls produced longer and had more completed lactations. It is not clear whether this was due to preference by farmers of their respective home strains or it had anything to do with production, health or fertility reasons. The frequencies of disposal reasons such as mastitis in Canada and pasture bloat in New Zealand were too low to have influenced the differences seen in lengths of productive life and number of lactations completed (Mwansa, 1997). Heritability estimates for most of the traits were higher in New Zealand than in Canada and estimates for first lactation yield traits were within estimates from literature. Heritability estimates for profit traits were in the medium range implying that it would be possible to select for economic efficiency. Genetic and phenotypic correlations between first lactation yield traits and economic efficiency traits in New Zealand were medium to high. On average fat showed higher correlations than milk or protein yields. Yield traits in the first lactation therefore appear to be a good tool in the selection for economic efficiency, either over partial or entire lifetimes. Correlation between economic efficiency over the first lactation and economic efficiency over productive life was moderately high. The ultimate selection goal is high economic efficiency over a cows productive life, which can either be selected for directly or indirectly. Since direct selection would mean longer generation intervals, correlated responses to first lactation traits would indicate the best traits to use in a selection scheme. Predicted correlated response of lifetime economic efficiency to  92  selection on first lactation protein was slightly higher than those from selecting for either milk or fat yield, but lower than that from selecting for economic efficiency over first lactation. This showed that the best trait that is measured early in production life to use in the selection for lifetime economic efficiency would be first lactation economic efficiency. The efficiency of this is only about 56% (1.8 vs. 3.2 $/l,000 Meal) relative to direct response, but this would be offset by the reduced generation interval.  93  3.7.  REFERENCES  Ahlborn, G. and Dempfle, L., 1992. Genetic parameters for milk production and body size in New Zealand Holstein-Friesian and Jersey. Livest. Prod. Sci. 31: 205-219. Allaire, F. R., 1981. Economic consequences of replacing cows with genetically improved heifers. J. Dairy Sci. 64: 1985-1995. Andrus, D. F. and McGilliard, L. D., 1975. Selection of dairy cattle for overall excellence. J. Dairy Sci. 58: 1876-1879. Balaine, D. S., Pearson, R. E. and Miller, R. H., 1981. Profit functions in dairy cattle and effect of measures of efficiency and prices. J. Dairy Sci. 64:87-95. Bar-Anan, R., Heiman, M., Ron, M. and Weller, J. I., 1987. Comparison of proven sires from five Holstein-Friesian strains in high-yield Israeli dairy herds. Livest. Prod. Sci. 17: 305322. Beaudry, T. F., Cassell, B. G., Norman, H. D. and Pearson, R. E., 1988. Impact of prices on profit functions in dairy cattle. J. Dairy Sci. 71:485-491. Bertrand, J. A., Berger, P. J., Freeman, A. E. and Kelley, D. H., 1985. Profitability in daughters of high versus average Holstein sires selected for milk yield of daughters. J. Dairy Sci. 68: 2287-2294. Blake, R. W., 1984. Considerations in multiple trait evaluation. J. Dairy Sci. 67: 1554-1566. Brascamp, E. W., Smith, C. and Guy, D. R., 1985. Derivation of economic weights from profit functions. Anim. Prod. 40: 175-180. British Columbia Milk Marketing Board, 1996. Annual Statistics, 1995-1996. Cassell, B. G., Pearson, R. E., Stoel, J. and Hiemstra, S., 1990. Relationship between sire evaluations for linear type traits and lifetime relative net income from grade or registered daughters. J. Dairy Sci. 73: 198-204. Cassell B. G. and Weigel, D. J., 1994. Use of lifetime net income in dairy genetics research. Proceed., 5th World Congress on Gene. Applied to Livest. Prod. Vol 17: pp 97. Dado, A. A., Mertens, D. R. and Shook, G. E., 1990. Theoretical metabolizable energy and protein requirements for milk component production. J. Dairy Sci. Suppl. 73:248. de Haan, M. H. A., Cassell, B. G., Pearson, R. E. and Smith, B. B., 1992. Relationships between net income, days of productive life, production and linear type traits in grade and registered Holsteins. J. Dairy Sci. 75: 3553-3561.  94  Dempfle, L., 1986. Increasing the efficiency of the dairy cow with regard to body size. Res. Bull. No. 4., Livestock Improvement Corporation Ltd. New Zealand Dairy Board. Hamilton, New Zealand. Dentine, M. R., McDaniel, B. T. and Norman, H. D., 1987. Comparison of culling rates, reasons for disposal, and yields for registered and grade Holstein cattle. J. Dairy Sci. 70: 26162622. Dickerson, G., 1970. Efficiency of animal production - molding the biological components. J. Anim. Sci. 30: 849-859. Dommerholt J. and Wilmink, J. B. M., 1986. Optional selection responses under varying milk prices and margins for milk production. Livest. Prod. Sci. 14: 109-121. Dunklee, J. S., Freeman, A. E. and Kelley, D. H., 1994. Comparison of Holsteins selected for high and average milk production. 1. Net income and production response to selection for milk. J. Dairy Sci. 77:1890-1896. Everett, R. W., Keown, J. F. and Clapp, E. E., 1976. Relationships among type, production, and stayability in Holstein cattle. J. Dairy Sci. 59:1505-1510. Falconer D. S., 1989. Introduction to quantitative genetics, 3rd Edition. Publ. John Wiley & Sons, Inc., New York, NY. Gill, G. S., and Allaire, F. R., 1976. Genetic and phenotypic parameters for a profit function and selection method for optimizing profit in dairy cattle. J. Dairy Sci. 59: 1325-1333. Gilmore, J. A., 197.7. The relationship of milk yield and other traits measured early on life to a dairy cattle profitability model including health and opportunity costs. Ph.D. thesis, North Carolina State Univ., Raleigh, N. C. Gibson, J. P., 1989a. Selection on the major components of milk: Alternative methods of deriving economic weights. J. Dairy Sci. 72: 3176-3189. Gibson, J. P., 1989b. The effect of pricing systems, economic weights, and population parameters on economic response to selection on milk components. J. Dairy Sci. 72: 3314-3326. Graham, N. J., Burnside, E. B., Gibson, J. P., Rapitta, A. E. and McBride, B. W., 1991. Comparison of daughters of Canadian and New Zealand Holstein sires for first lactation efficiency of production in relation to body size and condition. Can. J. Anim. Sci. 71: 293300. Harris, B. L., 1992. Linear programming applied to dairy cattle selection. PhD. thesis, Iowa State University, Ames. Harris, D. L., 1970. Breeding for efficiency in livestock production: Defining the economic objective. J. Anim. Sci. 30: 860-865.  95  Hietanen, H. and Ojala, M., 1995. Factors affecting body weight and its association with milk production traits in Finnish Ayrshire and Friesian cows. Acta Agr. Scand. 45: 17-25. Hocking, P. M., Mcallister, A. J., Wolynetz, M. S., Batra, T. R., Lee, A. J., Lin, C. T., Roy, G. L., Vesely, J. A., Wauthy, J. M. and Winter, K. A., 1988. Factors affecting length of herd life in purebred and crossbred dairy cattle. J. Dairy Sci. 71: 1011-1024. Hoque, M. and Hodges, J., 1981. Lifetime production and longevity of cows related to their sire's breeding values. J. Dairy Sci. 64:1598-1602. Jairath, L. K., Hayes, J. F. and Cue, R. I., 1995. Correlations between first lactation and lifetime performance traits of Canadian Holsteins. J. Dairy Sci. 78:438-448. Klassen, D. J., Monardes, H. G., Jairath, L., Cue, R. I. and Hayes, J. F., 1992. Genetic correlations between lifetime production and linearized type in Canadian Holsteins. J. Dairy Sci. 75:2272-2282. Lin C. Y., and Allaire, F. R., 1977. Relative efficiency of selection methods for profit in dairy cattle. J. Dairy Sci. 60: 1970-1978. Lin C. Y., and Allaire, F. R.., 1978. Efficiency of selection on milk yield to a fixed age. J. Dairy Sci. 61:489-496. MacAllister, A. J., Lee, A. J., Batra, T. L., Lin, C. Y., Roy, G. L., Vesely, J. A., Wauthy, J. M. and Winter, K. A., 1994. The influence of additive and non-additive gene action on lifetime yields and profitability of dairy cattle. J. Dairy Sci. 77: 2400-2414. Meyer, K., 1983. Scope for evaluating dairy sires using first and second lactation records. Livest. Prod. Sci. 10: 531-553. Meyer, K., 1985. Estimates of genetic parameters for milk and fat yield for first three lactations in British Friesian cows. Anim. Prod. 38:313-322. Meyer, K., 1991. Estimating variances and covariances for multivariate Animal Models by Restricted Maximum Likelihood. Genet., Sele., Evol. 23: 67-83. Mwansa, P. B., 1997. Genetic analysis of longevity in the Canadian and New Zealand dairy herds. PhD. thesis, University of British Columbia. National Research Council, 1989. Nutrient requirements of dairy cattle. 6th rev. ed. Update 1989. Natl. Acad, of Sci., Washington D. C. Norman, H. D., Cassell, B. G., Pearson, R. E. and Wiggans, G. R., 1981. Relation of first lactation production and conformation to lifetime performance in Jerseys. J. Dairy. Sci. 64: 104-113. Pearson, R. E. and Miller, H. E., 1981. Economic definition of total performance, breeding goals, and breeding values for dairy cattle. J. Dairy Sci. 64: 857-869.  96  Pearson, R. E., Vinson, W. E. and Meinert, T. R., 1986. The potential for increasing productivity through selection for increased milk yield and component yields. Proceedings of the 4th World Congress On Genetics Applied to Livestock Production, Edinburgh, Scotland, XIV: pp541. Peterson R. 1988., Comparison of Canadian and New Zealand sires in New Zealand for production, weight and conformation traits. Res. Bull. No. 5, Livestock Improvement Corporation Ltd. New Zealand Dairy Board. Newstead, N. Z. Peterson R., 1991. Evidence of a genotype/environment interaction between Canadian Holstein and New Zealand Friesian cattle under Canadian and New Zealand management systems. Proc. of the 42nd Annual Meeting of the European Association for Animal Production, Berlin, Sept. 9-12, Vol. 1: pp 49. Rendel, J. M., and Robertson, A., 1950. Some aspects of longevity in dairy cattle. Emp. J. Exp. Agric. 18:49. Renkema, J. A., and Stelwagen, J., 1979. Economic evaluation of replacement rates in dairy herds. I. - Reduction of replacement rates through improved health. Livest. Prod. Sci. 6: 15-27. Rogers, G. W., Hargrove, G. L., Cooper, J. B. and Barton, E. P., 1991. Relationships among survival and linear type traits in Jerseys. J. Dairy Sci. 74: 286-291. Rogers, G. W., Van Arendonk, J. A. M, and McDaniel, B. T., 1988. Influence of involuntary culling on optimum culling rates and annualized net income. J. Dairy Sci. 71: 3453-3462. Ruiz , F. J., 1991. Relationships among length of productive life, milk yield, and profitability of United States, Canadian, and Mexican Holstein sires in Mexico. Ph.D. thesis. Cornell Univ. Ithaca, NY. Sieber, M., Freeman, A. E. and Kelley, D. H., 1988. Relationships between body measurements, body weight, and productivity in Holstein dairy cows. J. Dairy Sci., 71: 3437-3445. Smith, C , 1978. The effect of inflation and form of investment on the estimated value of genetic improvement in farm livestock. Anim. Prod. 26: 101-110. Statistical Analysis System Institute, Inc., 1994. SAS Users Guide, Release 6.08. SAS Institute Inc., Cary, N. C. Stott, A. W. and DeLorenzo, M. A., 1988. Factors affecting profitability of Jersey and Holstein lactations. J. Dairy Sci. 71:2753-2766. Tigges, R. J.; Pearson, R. E. and Vinson, W. E., 1984. Use of dairy herd improvement variables to predict lifetime profitability. J. Dairy Sci. 67: 180-184. Van Arendonk, J. A. M., 1985. Use of profit equations to determine relative economic value of dairy cattle herd life from field data. J. Dairy Sci., 74: 1101-  97  Van Doormal, B. J., Burnside, E. B. and Scheaffer, L. R., 1986. An analysis of the relationship among stayability, production and type in Canadian milk recording programs. J. Dairy Sci. 69:510-517. Van Raden, P. M., 1988. Economic value of body size in Holsteins. J. Dairy Sci., 71 (Suppl. 1): 238 (abst.). Visscher, P. M. and Goddard, M. E., 1995. Genetic analysis of profit for Australian dairy cattle. Anim. Sci. 61:9-18. Vukasinovic, J. and Kunzi, Moll N., 1995. Genetic relationships among longevity, milk production, and type traits in Swiss Brown cattle. Livest. Prod. Sci. 41: 11-18. Weigel D. J., 1993. Relationships among estimated net income, herdlife and linear type traits in dairy cattle. Ph.D. thesis. Virginia Poly. Inst, and State Univ., Blacksburg, VA. Yerex, R. P., Young, C. W., Donker, J. D. and Marx, G. D., 1988. Effects of selection for body size on feed efficiency size of Holsteins. J. Dairy Sci., 71: 1355-1360.  98  CHAPTER 4: GENOTYPE BY ENVIRONMENT INTERACTION FOR ECONOMIC EFFICIENCY AND CONSTITUENT TRAITS  4.1. SUMMARY  The existence of genotype by environment interaction was tested for first lactation yield traits, estimated mature weight and economic efficiency traits of daughters of Canadian and New Zealand sires in Canadian and New Zealand herds. The test for the interaction in economic efficiency was done using British Columbia and Ontario milk payment systems. The interaction was tested at the macro level (strain by environment) and at the micro level (sires within strain of sire). Among the first lactation yield traits, significant genotype by environment interactions were found at the macro level for protein yield and percentage protein. Among economic efficiency traits a significant interaction was only realized for lifetime economic efficiency when using the British Columbia milk payment system for Canada. No significant interactions were seen for any of the other economic efficiency traits and none for mature liveweight. At the micro level significant interactions were realized for all economic efficiency traits among sires of both strains. Significant interactions were also realized for first lactation yield traits among Canadian bulls. This indicated that breeding values of sires from progeny performance in each of the environments changed ranks from one environment to the other. The implication was that the breeding values estimated in one environment cannot be used to predict the performance of the sires in the other environment for the traits with significant interactions.  99  4.2.  INTRODUCTION  The importation of germplasm, be it in the form of semen, embryos, or live animals, serves to broaden the existing genetic base. Regardless of this, however, the daughters of imported sires have to be competitive and indeed comparable to the daughters of native sires in their performance particularly in terms of their economic efficiency. Why particularly economic efficiency? Because economic efficiency can be viewed as a value trait, that is, a combination of many traits that gives the individual under assessment a holistic value which can then be perceived as a value of the individual's genotype. To justify the importation of germplasm, which in many cases will be more expensive than the local germplasm, the daughters should be equally, if not more, profitable than daughters of local sires. Sires, dams, and indeed individual cows are usually assessed on the basis of individual traits and sometimes based on an index combining several traits. Economic efficiency on the other hand is a composite value trait that is estimated based on other individual characters. Most studies on genotype by environment interaction (GxE) are based on individual traits which might show presence or absence of significant interaction. When a test for GxE is conducted, whatever the outcome is on the individual traits involved in the derivation of the value trait, no conclusion can be drawn regarding the existence of an interaction on the composite (value) trait. Namkoong (1985) tested for the existence of GxE in a value trait in the absence of the same in constituent individual traits and came to the conclusion that even in the absence of any changes in the genotypic ranking over environments on individual trait basis, it is possible for changes in ranking to occur when the composite trait of genotypic value is assessed over different environments.  100  The proposed study will try to examine the existence of genotype x environment interaction in individual traits and a value composite trait (economic efficiency) through the following objectives; (i) To investigate the existence of a genotype (strain) by environment (GxE) interaction in yield, weight and economic efficiency traits in the two countries. (ii) To investigate the existence of a genotype by environment interaction for sires within strain in yield, weight and economic efficiency traits and in the two countries.  4.3. REVIEW OF GENOTYPE BY ENVIRONMENT INTERACTION  Genotype (or strain) by environment interactions can be defined as a change in the ranking of genotypes/strains or in the magnitude of the difference in performance when two or more genotypes are tested in two or more different environments. This is best illustrated in Figure 4.1, in which Genotypes 2 and 3 would exhibit a GxE interaction due to change in ranking form Environment 1 to Environment 2. Genotypes 1 and 3 on the other hand would exhibit a GxE due to change in the magnitude of the difference between them. Genotypes 1 and 2 do not show any GxE since both ranking and the magnitude of the difference remain the same. The change in performance differences from one environment to another could be observed between strains/breeds or even between individuals of the same strains/breeds but different ancestries (lines). Thus in genotype by environment interaction studies, genotype is defined in terms of either strains, breeds or groups of breeds and crosses (Wang et al., 1992).  101  Two sires of the same strain might manifest changes in the difference between them in terms of the performance of their progeny in one environment as compared to another. Genotype by environment interactions could also be perceived from the differences in the variability of daughter performance when sires are tested in different levels of production performance. Lytton and Legates (1966) reviewed several studies and observed that when similar data have been stratified according to production levels, components of variance have been found to increase in magnitude as the production level increased. These changes in the magnitude of the genetic variances for the different environments represent a second manifestation of genotype by environment interaction. Robertson (1959) showed this to be related to the correlation between average breeding values in two different environments using the following equation,  102  ° GE 2  =  1 (o-  Gl  ~  <7  Cl  )  2  + 2<r  Gi  cr (l - r C2  GiG2  )  Hence, the genotype by environment (sire by environment) interaction is influenced by the difference in the genetic standard deviation in the two environments (o" and <J ) as well as Gi  the magnitude of the genetic correlation (r and T  GG  GG2  Gi  ). Interactions would not be present if o" = <y Gi  c  =1.0 (Lytton and Legates, 1966). When genotype by environment interactions exist at the individual sire level and they  affect important characters for example yield traits, economic efficiency and so on, it becomes complicated to compute an unbiased estimate of a sire's breeding value for these traits. Researchers have observed that ignoring interaction of sire and herd would lead to an overestimation of breeding values (Banos and Shook, 1990; Meyer, 1987). Van Tassell and Berger (1994) showed that even with relatively low levels of sire x herd (environment) interaction, biases in sire variance and heritability estimates were considerable when interaction was omitted from the model estimating the variance components. Lack of an interaction on the other hand implies that the results of sire testing have less bias than those obtained in situations where an interaction exists and that they are applicable across the environments in question. Thus a bull may be safely selected and used for mating cows in one environment, even though he was evaluated in the other environment. In this study the Canadian environment could be considered a high performance environment relative to New Zealand. The question then becomes whether differences in the levels of performance will translate to the existence of GxE as suggested by Lytton and Legates (1966).  103  4.3.1. Studies on genotype by environment interaction in dairy cattle  Numerous studies have been done on the existence of genotype by environment interaction for many traits of beef cattle (Sellers et al. ,,1970; Butts et al, 1971; Tess et al, 1979; Bolton et al, 1987). In dairy cattle however, information on the interaction of genotype by environment is a little less documented. Some studies have looked for the existence of GxE by considering various regions in North America (Lytton and Legates, 1966; Carabano et al, 1990), while others have looked at the levels of yield as a measure of the management or production level (De Veer and Van Vleck, 1990), others at Somatic Cell Count (Banos and Shook, 1990), and a few have considered different feeding regimes (Wiggans and Van Vleck, 1978; Hansen et al., 1982). A number of studies have looked at the existence of genotype by environment at the country level (Bar-Anan et al, 1987; Peterson, 1991; Metzger et al, 1994) where sires have been used in a foreign country as well as their native country. In a study looking at genotype by country interaction, Metzger et al. (1994) tested the existence of GxE in yield traits among U.S. and Danish Jerseys. They reported that correlations of official, national US and Danish EBV of the bulls in their study were high enough to provide little evidence of interaction of genotype and environment. They however did not report any results on strain by country interaction. Peterson (1991) (working with part of the same data used in the current study) tested for the existence of GxE for first lactation 305-day-standardized yield traits. He found no evidence of interaction between environment and strain in any of the traits, indicating the strains did not change ranks significantly from one environment to the other. Interaction between environments and sire within strain however, were significant for milk, fat and protein yields, suggesting that within strain, the sires ranked differently in New Zealand than in Canada.  104  Bar-Anan et al. (1987) also tested genotype x environment at the countries level for milk production by looking at the Predicted Differences (PDs) of sires of different strains when used in their native countries and in high-yield Israeli herds. They considered sire strains from United States, Canada, New Zealand, Sweden and Israel itself. The only strain that showed a significant genotype x environment interaction is the New Zealand strain, meaning the rankings of sires in the New Zealand and Israeli environments were significantly different. Stanton et al. (1991) also looked at the interaction of sire by country by considering genetic correlations for milk yield between breeding values in the US and in three Latin American countries. They reported that genetic correlations were relatively large suggesting that sires would rank similarly for ME milk in the two regions. They however observed that production resources are more limited and costly in Latin America resulting in lower profit margins. The study therefore recommended evaluation of sires in Latin America with a multi-trait index combining ME milk and other economically important characters such as stayability and reproductive traits. Even though milk yield did not show any changes in sire rankings, this might not be the case when multi-trait or composite traits are considered. A multi-trait trait might show a significant GxE interaction even where milk yield alone does not as was demonstrated by Namkoong (1985). Carabano et al. (1990) considered three US regions as representing three different environments for dairy production in the US (California, New York and Wisconsin) and used them to test for sire genotype x environment interaction. They reported that estimates of genetic correlations indicated no important GxE interaction for milk yield, fat yield or fat percentage among the regions examined. They therefore concluded that the sires with daughters in one region would not be expected to be significantly re-ranked on their progeny in another region of the US.  105  De Veer and Van Vleck (1987) in their study tested for the existence of GxE by considering different levels of production to be different environments. They however reported that correlations between sire values at different levels were close to unity, confirming previous evidence (Lytton and Legates, 1966) that ranking of sires is not affected greatly by production levels of the herds where daughters make their records. Rather than production level, Banos and Shook (1990) considered herds of different average milk somatic cell count (SCC) as different production environments. They also found no evidence of sire x herd interaction for milk yield among the herds of different average milk SCC. There are studies like those of Mao and Bumside (1969), Wiggans and Van Vleck (1978) and Wang et al. (1992) that have considered different dietary and feeding regimes to be different environments and looked for interaction between sire genotype and these dietary environments. Mao and Bumside (1969) reported a significant interaction of sire Predicted Transmitting Ability and amount of grain fed in a study of Canadian herds. Wiggans and Van Vleck (1978) on the other hand found results that did not suggest any interaction between sire genotype and the proportion of net energy from concentrates used in the herd feeding regime. Wang et al. (1992) worked with Canadian Holsteins and Ayrshires and examined the existence of genetic line by concentrate level (proportion of concentrate in diet) interactions. They reported that an interaction existed for milk production and feed efficiency, and that Holstein cows can produce more milk than Ayrshire cows with increasing amounts of concentrate. Sire by concentrate level interactions were also significant for milk production and feed efficiency. Feed efficiency could be considered a composite value trait since its computation involves the use of more than one trait and it is also possible to put monetary value to the inputs and outputs considered in the computation.  106  Mao and Burnside (1969) also tested for the existence of an interaction between sire proofs and the price received for milk. Their argument was that cows in herds with higher milk market prices generally tend to be better fed than those in herds receiving lower prices. The study showed that no significant interaction existed between the sire proofs and the milk price. Neither were there any significant interactions with source of water on pasture, forage feeding methods in winter and in summer, amount of exercise cows received in winter and herd size.  4.4. MATERIALS AND METHODS  The data used in the analyses for genotype by environment interactions (GxE) was as described in Chapters 2 and 3. The economic efficiency traits used here were economic efficiency up to the end of lactation one (PF,), and lifetime economic efficiency (PF ), with L  computations shown in section 3.3.2. Profitabilities in both environments were considered and in Canada economic efficiency was considered under both Ontario and B.C. pricing systems. Estimation for GxE was also done for total first lactation yields of milk fat and protein and for mature weight estimated using the Von Bertalanffy growth function. GxE interactions were considered at two levels, macro and micro. GxE at the macro level is the interaction of strain of sire and the environment. The interaction was tested by merging the datasets from the two environments and then running a mixed model which included the appropriate interaction term between sire strain and environment using the General Linear Model (GLM) procedure of SAS (SAS Institute Inc., 1990). The following model was used;  107  Yijkim = \i + Cj + Hj(Ci) + T + Ci*T +S,(T )+ Ci*S,(T ) + e k  k  k  k  ijklm  (4.1)  where; "Yjjklm  = an observation on the ijklm* cow,  Cj  = the fixed effect of the i* production environment (country)  Hj(Cj)  = the fixed effect of the j herd in the i* production environment,  T  = the fixed effect of the k* strain of sire(Canadian or New Zealand),  m  k  Cj*T  = the effect of the interaction between the production environment strain of  k  sire, Sj(T )  = the random effect of the 1 sire within strain k, th  k  Cj*Si(T ) = the effect of the interaction between the environment and sire within strain, k  ijklm  e  =  m  e  random error associated with the ijklmth observation.  The interaction effect between strain of sire and production environment (Tj*C ) in the model is k  a test for the existence of a GxE interaction at the macro level while the interaction between environment and sire within strain of sire will give an indication of the GxE at the micro level. The interaction Tj*C was declared significant if P < 0.05. k  The existence of a GxE interaction at the micro level is explained by changes in the ranking of sires on being assessed in more than one environment. Apart from the results obtained from the model above it was further tested byfirstpredicting the breeding values of the sires in each of the production environments (countries) using the animal model within environment given in Chapter 3 model (Eqn 3.20). The product moment correlations between the estimates of breeding values in the two environments were then tested against the expected correlations with the assumption that the traits were the same trait in both environments, and that, therefore, the genetic correlation between them was unity. The expected or theoretical  108  correlations were derived with this assumption of a genetic correlation of unity and were computed using the equation given by Calo et al. (1973). Bar-Anan et al. (1987) further developed the equation to look as follows;  E(r) = KRPTjj • RPTjiVdRPTu • lRPT ) 0  (4.2)  5  2i  where E(r) is the expected correlation between sire breeding values, RPTjj is the repeatability of the breeding value on the ith sire from progeny reared in environment 1 and RPT j is the 2  repeatability of the same sire from progeny in environment 2. The summation is done over the number of sires within a strain. The repeatability of breeding value (RPT), is the square of the correlation between the predicted and true breeding value (r ) referred to as accuracy of TI  evaluation. If the trait were the same under both environments, then the genetic correlation between the same trait measured in the two environments would be unity. Such an outcome would imply that the GxE is not different from zero. The r  TI  in this study were approximated  using the method described by Meyer (1989). Differences between expected and actual (product moment) correlations was tested using the method described by Steel and Torrie (1981). The two correlations are converted to Z's and tested as:  Z' = ^  E  1?J A  V(2/»-3)  where Z*~(0,1) is the calculated value of Z to be compared to the table value, Z and Z are the E  A  Z values of the expected and actual correlations respectively, and n is the number of sires.  109  4.5. RESULTS AND DISCUSSION  4.5.1. Genotype by environment interaction at the macro level  Significance for the strain x environment effect and other factors for the various traits are shown in Table 4.1. There was no significant strain x environment interaction for mature weight, for first lactation milk and fat yields and for percentage fat. This implies there was no evidence of GxE interaction for these traits at the macro level. For fat yield this was not surprising given that the differences between the two strains, within both environments were not significant. The means for the various traits in Canada and New Zealand are given in Tables 4.2 and 4.3, respectively. Even when the rankings did switch, for example in fat yield where New Zealand sired cows were better in Canada and Canadian sired cows better in New Zealand, it still did not make a significant difference because the changes involved were small and not significant. Similar results (lack of GxE at the macro level for milk and fat yields) were found by Peterson (1991) working with the same data but looking at first lactation yields of milk, fat and protein adjusted for days milked. The effect of strain x environment was significant for first lactation protein yield and percentage protein (P < .05).. The mean protein yield in Canada was 200 and 204 kilograms for daughters of Canadian and New Zealand bulls, respectively while in New Zealand the corresponding means were 120 and 116 kilograms, respectively. Cows sired by New Zealand sires produced milk with higher percentages of protein in both environments. Mean percentages in Canada were 3.23 and 3.27 among Canadian and New Zealand sired cows respectively while the corresponding percentages in New Zealand were 3.20 and 3.33 , respectively. The difference between strains for both yield and percentage of protein were significant (P < .05) in New Zealand but not in Canada. The manifestation of G x E in both protein traits is not due to a  110  * ft  00  o  o Cu ft  < * 00  00  i t  P  ft  o  %  §  3'  * w o <  a ft  <  w < ft  3  00  ft  'oo p  00  n P  5' o  5'  3'  ft r - f oo" g.  3"" ~ w ft  p  4^  O  OJ  to  OJ  oo  OJ  oo  OJ  A  O  S- o a- ui  oo ^1  so  N i U l  *  00 ^1  U l  OS  SO ©  ©  U l  ©  to  o  oo  4^ to  *  U l  +  o ©  +  o ©  +  © ©  +  4^ OJ  *  U l  ^) SO *  OJ  U l  *  SO  *  oo SO  *  Os *  © ©  +  TJ i-i o  U l  SO  a. Tj  4^  oo  ^. Bj  *  U l  © ©  +  to OJ  bo #  ield  ©  © ©  4^  to  B' 3  B[ Tj p  cf  N ° C X  TJ os  OJ  4^.  to  4^  os 4^  ON SO  4^  4^ #  U i  so U l  SO  4* #  00  SO OS  U l  #  © ©  +*  SO U l  bo  oo  OJ  *  © ©  +  © ©  +  o © +  o ©  o ©  © ©  * -o SO #  © ©  +  *  © ©  +  +  U l  U l  ©  © ©  + *  ©  +  © ©  © ©  +  *  00  bo * oo 00  *  bs  00 Os  SO *  4^ *  I—I ft  * 3'  © ©  +  #  Ov t—» *  Os  to #  Os SO *  © ©  +  U l  #  to #  TJ Tj  © ©  + © ©  +  00 ;-J  O  r-f  Os *  #  +  * so ps  + *  o © +  *  l-t  © ©  ©  +  TJ Tj  Table 4.2. Least squares means for first lactation yield traits, mature weight and economic efficiency traits for daughters of Canadian and New Zealand sires in Canadian herds Strain of sire Canadian  New Zealand  6254* (85)  6151 (94)  Fat yield  216 (3)  222 (4)  Protein yield  200 (2)  204 (3)  Fat %  3.50* (.04)  3.65 (.05)  Protein %  3.23 (.02)  3.27 (.03)  Milk yield  MtWt  652* (10)  628(12)  PFl(ON)  235 (1)  236(1)  PFl(BC)  264(1)  262 (2)  PFL(ON)  234 (1)  234 (1)  PFL(BQ  268* (1)  263 (2)  * differences between strain means significant at P < .0.05 (nn) standard error MtWt = mature weight  Table 4.3. Least squares means for first lactation yield traits, mature weight and economic efficiency traits for daughters of Canadian and New Zealand sires in New Zealand herds Strain of sire Canadian  New Zealand  Milk yield  3755* (85)  3491 (51)  Fat yield  151 (2)  149 (2)  Protein yield  120* (1)  116(2)  Fat %  4.08* (.03)  4.37 (.03)  Protein %  3.20 (.01)  3.33 (.01)  MtWt  429*(5)  416(5)  PF,  119(1)  120(1)  PF  120(1)  122(1)  L  * differences between strain means significant at P < .0.05 (nn) standard error MtWt = mature weight  switching of ranks by the strains in the two environments but rather due to a change in the magnitude of the difference between them.  A similar change in the magnitudes of the  differences was also seen in fat percentage which was 0.14 in the Canadian environment and 0.29 in New Zealand. This change did not, however, translate into a significant strain by environment interaction. The New Zealand environment with its predominantly pasture forage diet seems more conducive to the production of milk with higher percentages of fat and protein than the Canadian one. This could also be due to the genetic background of the dams since they are from a population mainly selected for fat production, and indeed of late both fat and protein. As earlier indicated the Canadian population has mostly been selected for milk production, at least up to the time the project sires were selected. The difference between two genotypes in the expression of their genetic potential for a given trait is generally better realized in a higher performance environment (with reference to the trait in question). In this case as far as percentages of fat and protein are concerned New Zealand could be considered the higher performance environment in that the percentages of fat and protein are higher there. The differences between the two strains were also better exemplified there. The GxE realized in these two traits could therefore mostly be associated with the larger differences seen in the "more conducive" New Zealand environment compared with the lack of differences in Canada. In looking at the effect of the strain x environment interaction for profit traits the two milk pricing systems for BC and Ontario were considered separately.  The profit traits  considered were economic efficiency up to the end of the first lactation (PF,) and economic efficiency over productive life (PF ). In Canada the mean economic efficiency up to the end of L  lactation one with Ontario prices was 235 $/l,000 Meal for Canadian sired cows and 236 $/l,000 Meal for New Zealand sired ones. The corresponding figures with BC prices were 264 and 262  114  $/1,000 Meal, respectively. In New Zealand mean economic efficiency over lactation one was 119 $/l,000 Meal for Canadian sired cows and 120 $/l,000 Meal for daughters of New Zealand bulls. Lifetime economic efficiency with Ontario prices was 234 $/l,000 Meal for daughters of both strains and 268 and 263 $/l,000 Meal with BC prices for Canadian and New Zealand sired cows, respectively. The corresponding figures in New Zealand were 120 and 122 $/l,000 Meal respectively. Interactions for both first lactation and lifetime economic efficiency using Ontario prices were not significant. With BC prices there was no significant interaction for profit up to the end of the first lactation but there was, however, a significant interaction for lifetime profit. This significant interaction of strain by environment with BC prices was probably due to the relatively larger difference between the cows sired by the two strains for lifetime economic efficiency under BC prices, which was significant (P < .05). The difference in New Zealand was, however, not significant. The implication of this significant interaction with BC prices is that milk pricing structure can introduce a GxE. In this study, economic efficiency (RET/FR) which is a value (composite) trait, did exhibited a strain x environment interaction, at least with the BC system of milk pricing. Among the traits involved in the computation the composite trait, namely milk, fat and protein yields and body weight, only protein yield displayed an important interaction. Thus in a way, the phenomenon postulated by Namkoong (1985) was observed in this study. That is, that even though individual traits might fail to show any GxE interaction, such an interaction might be exhibited by a value composite trait that is derived from a combination of several individual traits which individually do not exhibit any G x E . All components of variance for all traits were relatively much lower in the New Zealand environment than in the Canadian environment. As previously reported by Lytton and Legates (1966), with a higher level of production, the components of variance are also higher  115  4.5.2. Genotype by environment interaction at the micro level  The effect of sire within strain of sire as shown in Table 4.1 was significant for all profit, weight, and first lactation yield traits tested. The interaction of environment and sire within strain was also significant for all traits except protein percentage. This indicated that at least one sire caused a significant interaction due to either change in relative differences or change in rank. This necessitated a more comprehensive test utilizing the breeding values of the sires in the two environments using the method of Calo et al. (1973) as described in sub-section 4.3. The expected and product moment correlations between sire breeding values for both strains of sires are shown in Table 4.4. The expected correlations were computed based on the repeatabilities of breeding values of individual sires in each of the environments as described by Bar-Anan et al (1987). These expected correlations give the expected value of the correlations between the breeding values of the sires assuming that the two traits are the same in different environments and therefore have a genetic correlation of unity. Product moment correlations on the other hand are the actual correlations between the predicted breeding values in Canada and those in New Zealand. Among Canadian sires expected correlations ranged from .26 to .52, and product moment correlations from -0.18 to 0.40. The corresponding ranges among New Zealand sires were 0.24 to 0.48, and -0.26 to 0.43. All correlations were significantly different from zero. Product moment correlations in most of the traits among Canadian sires were negative, except for percentages of fat and protein. Only correlations for the profit traits were negative among New Zealand sires. Among Canadian sires expected E(r) and product-moment (R) correlations were significantly different for lactation one yields of milk, fat and protein and for all economic efficiency traits. The differences were not significant for mature weight, and percentages of fat and protein. Among New Zealand sires significant differences were realized for all economic  116  Table 4.4. Expected correlations [E(r)] and product moment correlations (R) of sire breeding values for weight, lactation one and economic efficiency traits. Canadian bulls New Zealand bulls E(r)  R  E(r)  R  Mature weight  0.55  0.45  0.50  0.19  Lactation 1 milk yield  0.29*  -0.18  0.25  0.39  Lactation 1 fat yield  0.34*  -0.04  0.31  0.34  Lactation 1 protein yield  0.33*  -0.08  0.29  0.43  Lactation 1 fat percent  0.52  0.40  0.48*  0.09  Lactation 1 protein percent  0.42  0.26  0.38  0.41  Lactation 1 profit (ON prices)  0.26  -0.03  0.24*  -0.13  Lactation 1 profit (BC prices)  0.30*  -0.18  0.27*  -0.15  Lifetime profit (ON prices)  0.38*  -0.04  0.35*  -0.26  Lifetime profit (BC prices)  0.33*  -0.07  0.30*  -0.12  * correlations (E(r) and R) significantly different (P < 0.05). All product moment correlations significantly different from 0 (P < 0.05).  117  efficiency traits and fat percentage but not for lactation one yield traits and protein percentage. A significant difference between the two correlations implies that sires of both strains change their rankings significantly when assessed in the alternative environment. This is evidence to suggest the existence of a GxE interaction at the micro level. Bar-Anan et al. (1987) also reported indications of a GxE in daughter milk yield of New Zealand sires when tested in both New Zealand and Israel. Working with the same data set used in this study, Peterson (1991) found evidence of GxE at the micro level for first lactation milk, fat and protein yields adjusted for length of lactation.  4.6. CONCLUSIONS  There was evidence of GxE interaction at the macro level for first lactation protein yield, percentage of protein and for lifetime economic efficiency using British Columbia prices in Canada. The significant interactions in yield and percentage of protein were due to the relatively large differences between the two strains in the New Zealand environment where Canadian sired cows produced more protein but at a significantly lower percentage. In Canada on the other hand there was no differences between the two strains. It was therefore due to a change in the magnitude of the difference between strains rather than a switching of ranks. The overall mean percentages of protein and fat were also higher in New Zealand than in Canada. The existence of a significant strain by environment interaction for lifetime economic efficiency with BC prices but not with Ontario prices implies that GxE can be as a result of not only the production environment itself but also of the pricing system. The lack of GxE using the Ontario pricing  118  system could be due to the similarity in the system and that used in New Zealand where they both pay based on protein and fat yields. The was no evidence of strain by environment (GxE) interaction in any of the remaining yield and profit traits and neither was there an interaction for mature weight. This means generally the differences between the two strains remained consistent across the two environments and even where the rankings changed, the change was not large enough to manifest an interaction. Thus even though they have basically been bred towards different breeding goals, the New Zealand strain for milk fat yield and the Canadian strain for milk yield, and under different production environments, this did not create a GxE at the macro level. There were differences seen in milk yield, protein yield and protein percentage, but no interactions for the former two. Most important to note however is that even though most of the traits did not exhibit a GxE the most important trait to the farmer, that is lifetime economic efficiency does so when the contrasting payment systems are the BC system in Canada and the New Zealand system. The three yield traits and economic efficiency traits displayed significant GxE interaction at the micro level among Canadian sires. This indicated sires within strains changed ranking when tested in the two environments. The implications are that the rankings of the sire proofs in their native countries can not be used in the foreign environment since the indication here is that they change significantly. Percentages of fat and protein and mature weight did not show a significant interaction implying sire proofs for these three traits in Canada can be adopted in New Zealand without having to test the bulls there. Contrary to the observation among Canadian bulls, the yield traits among New Zealand bulls did not show significant interactions at the micro level. This implies that sire rankings for these traits did not change much from one country to the other. No interaction was also seen for  119  percentage protein. Significant interactions were however seen for all economic efficiency traits and percentage fat. From these results it can be concluded that the relative performance of both Canadian and New Zealand bulls in the foreign environment for economic efficiency traits can only be predicted for sires that have daughters (having ties) in the two countries. The same can be concluded for Canadian bulls where the yield traits are concerned.  120  4.7. REFERENCES  Banos, G., and Shook, G. E., 1990. Genotype by environment interaction and genetic correlation among parities for somatic cell count and milk yield. J. Dairy. Sci. 73: 2563-2573. Bar-Anan, R., Heiman, M., Ron, M. and Weller, J. I., 1987. Comparison of proven sires from five Holstein-Friesian strains in high-yield Israeli dairy herds. Livest. Prod. Sci. 17: 305322. Bolton, R. C , Frahm, R. R., Castree, J. W. and Coleman, S. W.. 1987., Genotype x environment interactions involving proportion of Brahman breeding and season of birth. I. Calf growth to weaning. J. Anim. Sci. 65: 42-47. Butts, W. T., Koger, M., Pahnish, O. F., Bums, W. C. and Warwick, E. J., 1971. Performance of two lines of Hereford cattle in two environments. J. Anim. Sci. 33: 923-932. Calo, L. L., McDowell, R. E., VanVleck, L. D. and Miller, P. D., 1973. Genetic aspects of beef production among Holstein-Friesians pedigree selected for milk production. J. Anim. Sci. 37: 676-682. Carabano, M. J., Wade, K. M. and Van Vleck, L. D., 1990. Genotype by environment interactions for milk and fat production across regions of the United States. J. Dairy. Sci. 73: 173-180. De Veer, J. C. and Van Vleck, L. D., 1987. Genetic parameters for first lactation milk yields at three levels of herd production. J. Dairy. Sci. 70:1434-1441. Hansen, P. J., Baik, D. H., Rutledge, J. J. and Hauserm E. R., 1882. Genotype x environment interactions on reproductive traits of bovine females. II. Postpartum reproduction as influenced by genotype dietary regimen, level of milk production and parity. J. Anim. Sci. 55: 1458-1472. Lytton, V. H. and Legates, J. E.. 1966. Sire by region interaction for production traits in dairy cattle. J. Dairy. Sci. 49: 874-878. Mao, I. L. and Burnside, E. B., 1969. Sire by herd environment interaction for milk production. J. Dairy. Sci. 52: 1055-1062. Metzger, J. S., Hansen, L. B., Norman, H. D., Wolfe, C. W. and Pedersen, J., 1994. Comparison of United States and Danish strains of Jerseys for yield traits. J. Dairy. Sci. 77: 14571465. Meyer K. 1987., Estimates of variance due to sire x herd interactions and environmental covariances between paternal half-sibs for first lactation dairy production. Livest. Prod. Sci. 17: 95-115.  121  Meyer K. 1989., Approximate accuracy of genetic evaluation under an animal model. Livest. Prod. Sci., 21: 87-100. Namkoong G., 1985. The influence of composite traits on genotype by environment relations. Theor. Appl. Genet. 70:315-317. Peterson R., 1991. Evidence of a genotype/environment interaction between Canadian Holstein and New Zealand Friesian cattle under Canadian and New Zealand management systems. Proc. of the 42nd Annual Meeting of the European Association for Animal Production, Berlin, Sept. 9-12, Vol. 1: pp 49. Robertson A., 1959. The sampling variance of the genetic correlation coefficient. Biometrics, 15: 469. Sellers, H. I., Willham, R. L. and de Baca, R. C , 1970. Effect of certain factors on weaning weight of beef calves. J. Anim. Sci. 31: 5-12. Stanton T. S., Blake, R. W., Quaas, L. D., Van Vleck, L. D. and Carabano, M. J., 1991. Genotype by environment interaction for Holstein milk yield in Colombia, Mexico, and Puerto Rico. J. Dairy Sci. 74: 1700-1714. Statistical Analysis System Institute, Inc., 1994. SAS Users Guide, Release 6.08. SAS Institute Inc., Cary, N. C. Steel, R. G. D. and Torrie, J. H., 1981. Principles and Procedures of Statistics: A Biometrical Approach. 2nd Edition. Publ. McGraw-Hill. Tess, M. W., Kress, D. D., Burfening, P. J. and Friedrich, R. L., 1979. Sire by environment interactions in Simmental-sired calves. J. Anim. Sci. 49: 964-971. Van Tassell C. P. and Berger, P. J., 1994. Consideration of sire relationships for estimation of variance components with interaction of herd and sire. . J. Dairy Sci. 77: 313-324. Wang, S., G. Roy, L., Lee, A. J., Mcallister, A. J., Batra, T. R., Lin, C. Y., Vesely, J. A., Wauthy, J. M. and Winter, K. A., 1992. Genetic line x concentrate level interactions for milk production and feed efficiency in dairy cattle. Can. J. Anim. Sci. 72: 227-236. Wiggans, G. R. and Van Vleck, L. D., 1978. Evaluation of sires in herds feeding differing proportions of concentrates and roughages. J. Dairy Sci. 61:246-249.  122  CHAPTER 5: CONCLUDING REMARKS  The daughters of Canadian bulls weighed heavier at maturity than their New Zealand sired counterparts in both environments. In general the weight differences were larger in the New Zealand environment than in Canada. In New Zealand daughters of Canadian sires weighed significantly heavier from birth to maturity whereas in the Canadian environment the only difference in weight appeared at maturity. Bull calves were not compared in this study and thus no conclusion about them can be made. The New Zealand beef market mostly relies on dairy bull calves and culled dairy cows for its supply. It would therefore be interesting to compare the weights of bull calves bom of the two strains as this might have an economic impact. Daughters of Canadian bulls also produced significantly more milk in the first lactation in both environments and in New Zealand they produced more protein. Fat yields in both environments were similar but daughters of New Zealand bulls had significantly higher percentages of fat in both environments. This differences in production coupled with the disparities in weight did not appear to result in important differences in economic efficiency. To justify the importation of foreign sires (semen), those imported have to prove to be more profitable than the local ones. More often than not most imported semen, particularly from the top sires in their native countries tend to be relatively more expensive than semen from local sires. If such an investment does not translate to longer and more profitable productive lives from daughters, then it can not be justifiable. In this study, however, one cannot make a general conclusion at strain level since results showed a significant environment by sire within strain interaction effect in all the traits considered. It is possible that a foreign sire might be better than  123  some of the native sires in a given environment. The question of importation can therefore only be considered at the individual sire level rather than at the strain level. The daughters of the foreign strain in each of the environments also had shorter productive lives and fewer lactations completed than daughters of the local strain. Although no reason was found for this, it does further demonstrate that in general, the foreign strain do not show any advantage over the local one in either of the countries. First lactation yield traits were seen to be positively and moderately related to lifetime economic efficiency. Economic efficiency up to the end of the first lactation, however, was seen to be the best predictor of lifetime economic efficiency due to their highly positive (.73) genetic correlation. This could be partly attributed to the part-whole relationship between the two. In the New Zealand environment where genetic correlations were estimated, cows on average had 4.2 completed lactations and economic efficiency was seen to inrcrease with increasing number of lactations. The strong and positive relationship between economic efficiency up the end of lactation one and lifetime economic efficiency imply that we do not have to wait for daughter to be culled to progeny test a sire based on lifetime economic efficiency. As done in milk yield, first lactation records seem to be a good indicator of lifetime performance. The only traits that showed an important GxE interaction at the macro level were first lactation protein yield, percentage protein and lifetime economic efficiency with British Columbia prices. The fact that for the same trait with Ontario prices no GxE was seen underlines the importance of milk pricing system in the comparison of cow/strain economic efficiency. At the micro level all economic efficiency traits showed a significant interaction for both strains of sires, and first lactation yield traits for Canadian sires. For these traits breeding values obtained from progeny testing in each country cannot be transferred to the other country. Sires would have to be tested in each of the environments.  124  

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0088670/manifest

Comment

Related Items