A comparison of daughters of Artificial Insemination (AI) Holstein bulls demonstrated that the highest ranked bulls for production, health, or reproduction in confinement herds also ranked highest in grazing herds for the same trait. There is some genotype by environment interaction for production traits between confinement and grazing herds in the United States. Little re-ranking of bulls based on their daughter’s performance in the different systems was detected; and genetic progress can be achieved in grazing herds by selecting current active AI bulls for particular traits. However, the selection emphasis that different traits deserve economically in each system may be quite different.
Grazing as a form of low input dairy production to maximize profit is increasing in popularity in the United States. Herds where cows consume mostly grass forage produce, on average, less milk than their confinement. However, it has also been documented that the lower costs of production associated with grazing are more than enough to offset this decrease, and maintain or improve farm. A major concern for graziers is the choice of genetics for optimal performance in pasture-based systems. The primary question is whether a genotype by environment interaction (G×E) exists between the two distinct environments, confinement and grazing. In other words, does one expect those sires whose daughters are producing in both environments, to rank the same genetically in both environments?
Theoretically, such GxE differences exist and result from interactions between genotypes and environments. Under confinement systems, ideal cows produce the most milk with maximal energy inputs. Under sustainable pasture based systems, cows with optimal genotypes produce moderately high levels of milk given somewhat limited inputs. Even cows with the same genetic make-up may produce at different levels when exposed to different environments. Such differences in genetic expression are termed genotype-by-environment interactions.
Reproductive efficiency is important on any dairy, but especially to allow seasonal calving to match cyclic milk production and forage needs with seasonal availability of grass. Reproduction is influenced by level of production. Some graziers are turning to use of foreign genetics or natural service sires, due to a lack of satisfaction or confidence that U.S. genetics are best for their situation. This lack of confidence stems from genetic evaluations being dominated by records from confinement systems, in which production and reproduction breeding objectives may be somewhat different.
Accurate genetic evaluations of animals on a national basis require consistent sire rankings across all environments. Differences in the ranking of sires across different herds, regions, or management systems indicate the presence of interactions between genotype and environment. Understanding these interactions is vital for selecting the best sires for different management systems.
Compelling evidence supports the existence of genotype-by-environment interactions for cattle managed in conventional versus pasture-based systems. First are the genetic correlations between countries that participate in the Interbull international sire evaluation (Banos, 1997). The magnitudes of these genetic correlations indicate the similarity between the sire rankings across countries. Correlations near 1.0 indicate that little or no interactions between genotype and environment exist. However, as correlations decrease, the potential of sire re-ranking increases. The correlation between the United States and Canada is 0.96, indicating that daughters of the same sire will generally perform similarly in each country. Not coincidentally, the production systems in both nations are very similar. Most producers rely on confining cows and providing primarily stored feeds. In fact, dairy producers in most countries around the world also feed their cattle primarily with stored feed.
For most countries, the genetic correlations with the United States and Canada are about 0.90 or greater for milk yield. The obvious exceptions are for New Zealand and Australia. For these countries, the genetic correlations with the US are only 0.76 and 0.81, respectively (Wickham and Banos, 1998). Not surprisingly, dairy production in these areas is primarily pasture-based and widely different from the norm in North America. Moreover, there has been widespread usage of North American dairy bull semen for genetic improvement in Australia and New Zealand, such that low genetic correlations are not likely to arise strictly from divergent gene pools.
Within Australia, evidence of genotype-by-environment interaction has also been reported. Fulkerson (1997) compared the production of cows from genetic lines selected for high and average production when fed only pasture, pasture and a low level of stored feeds, and pasture and a high level of stored feeds. He found that the high genetic line produced more milk than the average line did under all three diets. However, the difference between lines decreased as less stored feed was fed. Cows from the high line also tended to lose more weight during lactation than did the cows from the average line when fed the pasture-based diet. A similar study in Ireland by Dillon et al. (1998) found no significant effects of genotype-by-environment interaction, but the differences among treatment groups in amount of concentrates fed were much less than in the Australian study. Another Irish study by Cromie et al. (1998) compared the genetic correlations among the top and bottom Irish herds for amount of feed concentrate fed per cow. They reported correlations of 0.89 for fat yield, 0.91 for protein yield, and 0.92 for milk yield. Correlations in this range are high, but can result in substantial re-ranking of sire genetic evaluations.
Recently, Cienfuegos-Rivas et al. (1998) estimated the genetic correlation of milk yield between the Northeastern United States and Mexico, searching for evidence of genotype-by-environment interaction. Although management intensive grazing is not a defining feature of the dairy industry in Mexico, one can generally describe the management of dairy herds in Mexico to have lower inputs and to be less intensively managed than are conventional United States herds. The genetic correlation between the U.S. and Mexico was only 0.63, indicating a significant level of interaction among genotypes in the two different environments. Stanton et al. (1991) found a higher genetic correlation between US and Mexican milk yield (0.90), but reported a low correlation (0.78) between U.S. and Colombian milk yield.
Recent studies in the US have examined the effects of genotype-by-environment interactions on milk production (Dimov et al., 1995 and 1996) and somatic cell score (Dimov et al., 1995). These studies examined the effects of the interactions by estimating the proportion of phenotypic variance in each trait accounted for by interaction effects of sires across different herds. Results indicated that less than 5% of the phenotypic variance in the respective traits was accounted for by interactions of sire and herd. However, grazing and confinement herds were pooled in their study and gross effects of such divergent environments may be important.
For the national genetic evaluation, the USDA currently assumes that 14% of the variance in production traits is associated with interaction of sires and herds (Wiggans and VanRaden, 1989). The use of such an inflated value is intentional, to help guard against the bias associated with preferential treatment for sires with many daughters in a single herd. Dimov et al. (1995) concluded that the differences in accuracy of sire estimated breeding values associated with using a realistic rather than inflated value for the variance of interaction effects were minor.
Weigel and Pohlmann (1998) identified 27 herds in Wisconsin in which at least 50% of annual feed intake was obtained from pastures. They used DHI records to compare the performance in these herds of daughters of commercially available sires to the performance of cows in conventionally managed herds. They reported that phenotypically, milk yield was similar in both environments, but that milk fat percentage was lower in the pastured herds. They attributed this to the difficulties faced by graziers in maintaining consistent quality of pasture throughout the growing season. When examining genetic effects, they found that sire PTA’s predicted daughter performance in pasture and confinement systems equally well for milk and protein, but not for fat. Because of the limited scope of their study, they recommended that similar work be repeated using more herds, and thus allowing comparison of daughters of more sires with offspring in both management systems (1997, personal communication).
Researchers at the Scottish Agricultural College (Veerkamp, et al., 1995) found similar results when comparing the performance of selected and control lines of dairy cattle when fed two different total mixed rations (neither including pasture) with either high or low levels of concentrate. Phenotypically, no significant effects of interaction were found. However, the regression of actual performance on pedigree index was slightly lower for the cows on the low concentrate diet than on the high concentrate diet. As was the case in the Wisconsin study, fat production was the trait most strongly affected. The genetic correlation between fat yields on the low versus high concentrate diets was only 0.64. That study only considered the effects of different feeds but not overall management system.
These results might seem to suggest that effects of changes in sire rankings between conventional and pasture-based systems are rather small. However, these studies were not all designed to estimate the differences in ranking of sires across conventional and pasture-based systems. When data was taken randomly from national databases with no concern for the specific production system employed for each herd, it is likely that records from pasture-based operations comprised a relatively small proportion of the total records. Also, herds were not grouped according to production system, so any differences in effects common to specific types of herds would have been difficult, if not impossible, to detect. A study such as the one proposed here is needed to accurately determine if effects of genotype-by-environment interaction exist for conventional versus pasture based systems.
Furthermore, all of the studies cited have focused on the traits associated with milk production. Although sales of milk are the primary source of income for dairy producers and milk production is, therefore, the most important genetic trait for all dairy producers; other traits have considerable genetic and economic value. Reproductive traits are among these other traits. Average conception rates have decreased from 66% to less than 50% in the US since the 1950s (Butler and Smith, 1989) despite advances in veterinary technologies. This decrease in conception rate has coincided with substantial genetic gains in average production per cow and, therefore, increased selection for milk production has been implicated in part in this decrease in conception rate. The genetic correlation between production and reproduction is generally in the range of about -0.20. In particular, recent studies have demonstrated that high producing cows tend to have delayed first observed estrus (Harrison et al., 1990, and Senatore et al., 1996). The precise physiological reasons for these effects are not well understood. However, the period of negative energy balance that high producing cows endure in early lactation when their energy intake does not match the energy produced in their milk could evoke a number of consequences that have a detrimental effect on reproductive function (Webb et al., 1998), thus delaying estrus.
Little is known about the effects of genotype-by-environment interaction on reproduction. Although, reproductive traits are clearly of considerable importance to all dairy producers, they are of particular concern to graziers and their interactions with production may be more important. First, many graziers practice seasonal calving to match times of greatest milk production to seasons of greatest pasture yield. In such herds, cows must conceive within a window of 60 days or less, because otherwise the subsequent calving will occur too late in the next growing season. Zartman and Shoemaker (1994) pointed out additional steps necessary on an intensive grazing dairy system to ensure cows successfully conceived in a short window of opportunity. These measures included rather expensive use of weekly veterinary support, milk progesterone assays, and hormonal drug therapy. Routine practices of heat detection and chalking of rumps was intensified. Dairy producers in conventional management systems are not faced with this restriction of a narrow window of opportunity to successfully breed cows.
A second reason reproduction may be more important for pasture-based systems is that even the highest quality pastures may lack the concentration of energy needed to meet the requirements of a high producing dairy cow, increasing the potential for negative energy balance. The relatively high ratio of protein to energy in fresh pasture may exacerbate the problem by increasing the concentrations of nitrogen in the blood, further compromising reproductive efficiency. These effects would effectively increase the magnitude of the antagonistic genetic correlation between production and reproduction in pasture-based herds, requiring an increase in the relative weight on reproduction versus production in the selection program.
The overall aim of this proposed project is to determine genotype-by-environment interactions that may be important when selecting among US dairy sires for the ability of their daughters to produce milk and reproduce efficiently in sustainable pasture-based versus confinement systems. Different genetic profiles may be optimal under each situation. Genotype-by-environment interactions fall into two categories, changes of scaling and changes of ranking. Changes in scaling, often simply reflecting different production levels, can be handled quite easily in genetic evaluation procedures. While changes of scaling affect the magnitude of breeding values (predictors of the genetic transmitting ability of a cow or bull based on its own records and those of its relatives), they do not usually affect the rank order of genetic evaluations. Thus the impact on which animals are chosen in separate environments is minimal. Changes in ranking cause the rank order of breeding values to differ depending upon the environment in which the cows milk weights or other records were produced.
Specific objectives and their rationale were:
1. To test for genotype-by-environment interactions among dairy cows managed under pasture-based versus confinement systems.
Genetic selection is an inexpensive means for dairy producers to make permanent improvements in the average productivity and efficiency of their herds. The Animal Improvement Programs Laboratory (AIPL) of USDA systematically evaluates the dairy cattle population and predicts the breeding values of production traits for all bulls and cows on approved milk recording programs. Rapid genetic improvement for milk, fat, and protein yield has been achieved with the current system, but yield records of daughters producing milk in confinement systems dominate results.
Dairy cattle that perform best while being fed stored feed with high concentrate to forage ratios in confinement may not be best suited for a pasture based system. Indeed, milk production may be slightly different traits genetically under each system. Different genes may affect milk production in each system or the same genes may have different sized effects across production systems.
In the US, little information is available to assist producers in making selection decisions based on cows’ genetic ability to reproduce. The genetic ability of cows to reproduce is especially important for seasonal calving, which is a common practice in sustainable grass-based dairy production. Thus, current systems may not identify sires of cattle that perform most efficiently in pasture-based systems. On the other hand, if GxE interactions are not important, then all producers can use the results of the USDA evaluation with confidence. If major effects of interaction exist among environments or traits, then changes to the current evaluation system may be warranted to ensure that cattle, especially bulls used for artificial insemination, can be ranked properly for both conventional and pasture-based systems.
2. To identify differences in the management factors of conventional and pasture-based systems that contribute to genotype-by-environmental interactions.
Confinement dairies usually attempt to maximize profit by maximizing milk production per cow, whereas pasture-based producers are more likely to maximize profit by reducing input costs, while maintaining adequate production. Management factors are likely to differ between the types of operations as certain input and production parameters receive varying levels of emphasis. Confinement and pasture-based herds may differ in factors such as average milk yield, breeding practices, lactation length, age at calving, and cull ratio, to name a few. Degree of variability in these traits may also differ. Accounting for these differences in the statistical models now used for genetic evaluation may help to eliminate the effects of genotype-by-environmental interactions and provide accurate ranking of genotypes under all management systems.
Alternatively, animals may rank the same under both management systems, but the relative economic importance of certain genetic traits may differ under each system. For example, reproductive efficiency of cows and fertility of bulls may be especially important when seasonal calving is practiced. This study will help identify which traits are most important for pasture-based systems of production and provide valuable information to help determine selection goals that are specific to such systems.
3. To educate pasture-based dairy producers about the implications of genotype-by-environment interactions on their sire selection goals and decisions.
Managing cattle and maintaining profitability on a truly low input and sustainable dairy grazing system is complex. The complexity is compounded when the grazier attempts seasonal calving to match grass production to milk production cycles. The only genetic information currently available for US dairy sires is primarily from data originating from confinement dairy production systems. Graziers need to know whether those genetic evaluations of dairy bulls serve as accurate predictors of how the offspring of those bulls will perform under grazing conditions. Hence, they will be the primary target audience for results of this study.
If we find no evidence for genotype-by-environment interactions, graziers will be able to select the best available US genetics, regardless of source, and be confident the resulting heifers will do well. On the other hand, if genotype-by-environment interactions exist, graziers will want to know which bulls perform best under grazing conditions and where to find optimal genetics, either semen or natural service bulls. Providing information to graziers through popular press articles and grazing conferences and by packaging and providing educational materials resulting from this research to dairy professionals who work with graziers will allow them to make wise genetic selection decisions.
The first and perhaps most critical step in this study was to identify a group of herds in Canada that practiced intensive grazing. A list of potential herds was identified based on contacts with dairy extension personnel, university professors, milk recording supervisors, semen salespeople, and membership lists of grazing interest groups. A questionnaire was drawn up to ask breeders about their feeding, general management, and breeding practices and was sent to 69 different herd owners. Based on the responses to this questionnaire, 22 herds were selected for the grazing group.
These herds were required to intensively graze their cattle for at least six months per year, during which pasture provided at least 70 percent of the forage in the cows’ diets. In addition, herds were required to be enrolled in DHI and to record pedigree data. All herds were in either Nova Scotia or Ontario. A control (confinement) group consisted of 34 large (>100 cows) free-stall operations in the same two provinces. Effects of G E were studied for three types of traits: 1) production, 2) reproduction, and 3) conformation.
Production. For production, traits of interest were yields of milk, fat, and protein and percentages of fat and protein. Data were 305-day mature equivalent (305ME) DHI records from 1990 and later and included 6,749 lactations from 2,817 cows in the grazing herds and 29,371 lactations from 12,774 cows in the control herds.
To investigate G E, two approaches were taken for production traits. The first approach tested for a scaling effect on yield traits, such as that observed by Veerkamp et al.(1995), in which the phenotypic differences in production of cows of differing genetic values varied according to the environment. To test for such a scaling effect, herds were separated into grazing and conventional groups and lactation records of cows in each group were regressed on the PTA of their respective sires. Significant differences between environments in the regression coefficients for PTA were assumed to indicate a scaling effect associated with the different environments. Weigel et al. (1999) applied a similar approach to production records from grazing and conventional herds in Wisconsin.
The model equation for this analysis was:
yijk = HYSi + PAj + PTAk + eijk, 
where, yijk = 305ME record for milk, fat, or protein yields of cow k, calving at at age j, in herd-year season i; HYSi = fixed effect of ith herd-year-season i (4 seasons per year); PAj = fixed effect of jth age and parity at calving (4 age groups within 3 parity classes: 1, 2, >3); coefficient of regression of the production record on the PTA of the sire of cow k; and eijk = random residual. In theory, is expected to be 1, under the assumption that the PTA represents the true transmitting ability of each bull. The PTA were ½ times EBV from the May 2001 Canadian genetic evaluation.
A second analysis of the data was undertaken to examine if yields of milk, fat, and protein and percentages of fat and protein were actually different genetic traits in the grazing and conventional environments. This hypothesis was analyzed by estimating the genetic correlation between records for each trait expressed under the two different environmental conditions. Correlations significantly less than 1.0 were assumed to indicate G E and statistically significant correlations <0.80 were assumed to indicate biologically important G E. Therefore, each cow had the information reported for a given trait in one environment and missing for the other. The genetic relationships from common sires were used to obtain the correlations.
A bivariate animal model with repeated records was assumed to describe the data. The model equation for this analysis was,
yijkt = HYSPi + PAj + Akt + PEk + eijk, 
where, yijkt = 305ME milk, fat, or protein, or fat or protein percentage for cow k in environment t, herd-year-season-parity group i and parity-age group j; HYSPi = fixed effect of herd-year-season-parity (first or later) i; PAj = the fixed effect of the parity-age of calving group j (4 age groups in 3 parity groups, first, second, and later); Akt = the random additive genetic effect of cow k in environment t; PEk = the random permanent environmental effect of cow k; and eijk = random residual. Variance components were estimated by REML using the VCE 4.0 software of Groeneveld and Garcia-Cortes (1998).
Reproduction. For reproduction, effects of G E were examined for calving interval (CI). The CI for a given lactation was simply recorded as the difference in days from the date of calving for that lactation and the following date of parturition. Because some cows did not have a subsequent calving, the data set included fewer records than had been available for production. In addition, CI was required to be >310 d, to avoid the inclusion of data from abortions or otherwise abnormally short pregnancies. This analysis included 3887 records (1784 cows) from the grazing herds and 16,461 records (8091 cows) from the conventional herds. For this trait, only the genetic correlation between environments was of interest and regression on sire PTA was not possible because no genetic evaluation for CI is performed in Canada. The model applied was the same as that used for production traits (Equation ). In addition, the genetic parameters were re-estimated using a model that considered phenotypic milk yield (305ME) as a covariate.
Conformation. Approximately 30 different conformation traits are recorded and evaluated in Canada, but many of these traits are correlated with each other and have relatively little economic importance. Therefore, we chose to concentrate our evaluation to the three traits that contribute directly to the Lifetime Profit Index, the official national selection index in Canada (Canadian Dairy Network, 2001). These traits are Mammary System, Feet and Legs, and Frame and Capacity.
Not all of the cows with production information had been classified, and only first lactation records were used because classification in later lactations is not obligatory (i.e. data from later lactations are likely to be a biased sample) and only first parity records contribute to the national genetic evaluation in Canada. For these reasons, the number of records for conformation were less than were available for production or reproduction. A total of 11,963 records were used, 2,191 from the grazing herds and 9,772 from the conventional herds.
Following the approach used for production and reproduction, genetic correlations were calculated between the same trait recorded in the two different environments.
The model equation for conformation traits was,
yijklt = SAi + RCj + HRCk + Alt + eijkl, 
where yijklt = conformation score of cow l in environment t, in age and stage of lactation group j, round-classifier group j, and herd-round classifier group k; SAi = fixed effect of age and stage of lactation group j (4 age and 11 stage groups); RCj = fixed effect of round-classifier j; HRCk = the random effect of herd-round-classifier k; (4 age groups in 3 parity groups, first, second, and later); Alt = the random additive genetic effect of cow l in environment t; and eijkl = random residual.
Grazing herds were defined as those in which cows consumed the majority of their forage from pasture for at least six months of the year. These herds were identified with the help of DHIA personnel, dairy specialists, and grazing networks. Herds had to be enrolled in a DHI recording program. Data were provided by Dairy Records Management Services, (Raleigh, NC) and AgSource, Inc. (Verona, WI). In addition, DHI herds of similar size from the same states but not known to be utilizing grazing were selected as control herds. In some states, few control herds were available. Therefore more control herds from other states were requested to balance the data set.
The grazing data consisted of 82,479 records of 34,566 cows in 393 herds in 12 Eastern US states, while the control data consisted of 142,924 records of 60,667 cows in 432 herds in 13 Eastern US states (Table 1). A requirement of regression analysis was that cows have an AI sire with USDA-DHIA Predicted Transmitting Abilities (PTA) for each trait. Three separate regression analyses were preformed for first lactation records of cows. For milk, fat, and protein analyses, regression coefficients were estimated for all records, for four random subsets, and for four subsets based on quartiles of herd average MEM. Common edits to the data set were: 1) only records after 1990 were included, 2) cows were required to have a first lactation, 3) cows were required to have pedigree information, 4) parities greater than 5 were deleted, 5) cows with records ± 4 phenotypic standard deviations of mean mature equivalent milk (MEM), fat (MEF), and protein (MEP) yield were deleted, and 6) cows with <60 days in milk for each group were deleted. The USDA Animal Improvement Programs Laboratory provided complete pedigree information going back three generations.
The performance of daughters in grazing and confinement was examined using linear regression of MEM, MEF, and MEP on the November 2000 USDA-DHIA PTA of their sires for those traits using the following model:
Yijkl = hi + ysj + apk + PTAijkl + eijkl, 
where Yijklm is mature equivalent milk, fat, and protein yield for the first lactation record of cow l, in herd i, calving in year-season j, in age-parity class k; hi is fixed effect of herd i; ysj is fixed effect of the year-season of calving j; apk is fixed effect of age-parity class k; is coefficient of regression of daughter lactation yield on sire PTA; PTAijkl is November 2000 USDA-DHIA PTA for milk, fat, protein, of the sire of cow l; and eijklm is random residual.
The univariate regression model was fit separately to grazing and control data. Quartile analysis was performed to observe the adequacy of prediction of the sires’ PTA at different levels of production.
Restricted Maximum Likelihood using the (co)-variance component estimation program, VCE4 (Neumaier and Groeneveld, 1998), was used to estimate heritabilities and genetic correlations between the two environments for each trait. The REML algorithm chosen for this study used analytical gradients. A bivariate animal model that considered traits in different environments as separate traits was used to estimate the variance components. Due to the large computational demands, heritabilities and genetic correlations were estimated for four random subsets of the overall data set. Overall estimates presented in this paper, are the mean of estimates for each of the four subsets, weighted by the number of observations in each subset.
To determine if there was re-ranking between the upper and lower producing herds a similar analysis was conducted for milk, fat, and protein, between the upper and lower quartiles and between the middle quartiles defined by herd average MEM, within each environment.
To assess the level of re-ranking among sires within each environment, product-moment and rank correlations between EBV were calculated using PROC CORR (SAS, 2000). Correlations were calculated for sires that had at least five daughters producing in both grazing and confinement herds. Predictions of breeding values were obtained for all sires separately in both the grazing and control groups using PEST (Groeneveld and Kovac, 1990). Correlations were also calculated between the sires’ EBV in either grazing or control herds with their November 2000, USDA-DHIA PTA. The following model was used for the estimation of the genetic parameters and estimation of the breeding values:
Yijkl = hysi + apj + ak + pek + eijkl , 
where Yijkl is mature equivalent milk, fat, protein yield for the lth record of animal k, calving in herd-year-season i, in age-parity class j; hysi is fixed effect of herd-year-season of calving i; apj is fixed effect of age-parity class j; ak is random additve genetic effect of animal k; pek is random effect of permanent environment for animal k; and eijkl is random residual.
Health and reproduction. Linear somatic cell scores (LSCS) was used as a measure of mastitis resistance. The number of days open (DO), days to first service (DFS) and the number of service per conception (SPC) were used as indicators of reproductive performance. Records for LSCS, DFS, SPC were available for DRMS herds only, while DO were available from both processing centers. A requirement of the regression analysis was that cows have an AI sire with USDA-DHIA Predicted Transmitting Abilities (PTA) for SCS. The performance of daughters in grazing and confinement for LSCS was examined using linear regression of daughter LSCS on the November 2000 USDA-DHIA PTA of their sires for SCS used model .
The univariate regression model was fit separately to grazing and control data. Regression coefficients were estimated for all first lactation records and for four random subsets for LSCS.
Methods of analysis were similar to those for production traits. The following model was used for the estimation of the genetic parameters and breeding values:
Yijkl = hysi + apj + ( tdm2l)+ ak + pek + eijkl , 
where Yijkl is lactation SCS or fertility measure for animal k, calving in herd-year-season i; in age-parity class j; hysi is fixed effect of herd-year-season of calving i; apj is fixed effect of age-parity class j; tdm2l is regression of trait on second test-day milk yield (for DFS and SPC); ak is random additve genetic effect of animal k; pek is random effect of permanent environment for animal k; and eijkl is random residual.
[Tables cited are in the Appendices]
Objectives 1 and 2
Results. Table C.1 has means and standard deviations for the different traits for both environments. Not surprisingly, average production was greater in the conventionally managed herds than in the grazing herds (P < 0.01). The difference in milk yield between groups was 547 kg, which was less than that (1259 kg) observed by Kearney et al. (Appendix E) for herds across the Eastern US, but greater than that (370 kg) reported by Weigel et al. (1999) for herds only in Wisconsin. Fat and protein percentages were very similar across environments. No significant difference was observed for CI. Kearney et al. (Appendix E) found that grazing herds in their study had slightly longer days open than did the control herds (154 to 148 d), but contributed this difference to climatic effects as a greater proportion of their grazing herds were from the southern United States. Among the type traits, the mean scores for mammary system and feet and legs were slightly but significantly greater (P < 0.01) within the grazing herds. The slight difference in udders may have been due to the slightly lower level of production in the grazing herds, as milk yield phenotypically is unfavorably correlated with traits such as fore udder attachment and udder depth (Short and Lawlor, 1992). The environment provided by grazing herds may have been favorable for feet and legs, as increased exposure to concrete has been shown to be associated with locmotive diorders (Galindo and Broom, 1993). Table C.2 has the regression coefficients of phenotypic yields in the different environments on sire PTA. For all three traits, the regression coefficients were greater (P < 0.01) for the conventional herds, indicating a scaling effect, and suggesting that genetic differences among animals were more obvious in conventional herds. For the conventional herds, the regression coefficients were all significantly (P < 0.01) greater than 1.0. For the grazing herds, all coefficients were less than 1 but this difference was significant (P < 0.05) only for fat yield. Weigel et al. (1999) reported similar findings for fat yield among grazing herds in Wisconsin. Kearney et al. (Appendix E) reported regression coefficients less than 1 for all three yield traits in their grazing herds. Regression for PTA were closer to 1 within the grazing herds, suggesting that the current EBV for sires in Canada predict differences among daughters more accurately in grazing herds than in conventional herds, at least among herds defined as conventional in our study. Veerkamp et al. (1995) found similar results when comparing the performance of selected and control lines of dairy cattle when fed two different total mixed rations (neither including pasture) with either high or low levels of concentrate. The regression of actual performance on pedigree index was slightly lower for the cows on the low concentrate diet than on the high concentrate diet. Estimates of heritabilities within each environment and genetic correlations across environments are in Table C.3 for all nine traits evaluated. Heritabilities were generally lower the grazing herds, but these differences were non-significant for most of the traits. For the yield traits, heritabilities fell in the general range of 0.35 to 0.40 for the conventional herds versus only 0.30 to 0.35 for the grazing herds. The increased estimates of heritability in the conventional herds coincide with the trends observed in regression coefficients (Table C.2), which suggested that genetic differences among animals were expressed to a greater degree in the conventionally managed herds. Heritiabilities of percentage traits were quite high but consistent with other estimates previously made on the Canadian Holstein population (Boettcher and Gibson, 1997 and Z. Lui, 1996, unpublished technical report). Effects of G E on yield traits, as indicated by genetic correlations, were minor. The estimates of genetic correlations between the two management systems were around 0.90 or greater (Table C.3) well above the threshold of 0.80 suggested by Robertson (1959). The genetic correlation was significantly (P < 0.01) less than unity only for fat yield. Fat yield also had the lowest genetic correlation in the study of Weigel et al. (1999), but was not different from 1. They attributed apparent interactions for fat yield to inconsistency in pasture quality during summer months in Wisconsin. Average correlations reported by Kearney et al. (Appendix E) were similar for all three yield traits, falling between 0.88 (fat) and 0.91 (protein). Correlations for percentage traits were different from 1 (P < 0.05), but >0.92.
Even less evidence of G E was observed for reproduction and conformation traits. Some indication for G E was observed for CI, as estimated heritability in grazing herds (0.052) was nearly double that in conventional herds (0.027) and estimated genetic correlation across groups was only 0.64. However, these differences were not significant (P > 0.05). When milk yield was added to the model as a covariate, results changed very little except that the estimate of heritability in the grazing herds decreased to 0.036.
Among the type traits, heritability for udder traits was greater in the conventional herds (0.20 vs. 0.13) but genetic correlations between groups was 1.00. Perhaps the greater heritability in conventional herds was associated with higher mean production, and thus more stress on udder attachments. For Feet and Legs and Frame and Capacity, heritabilities were not different across environments and genetic correlations were not different from unity.
Discussion and conclusions. In general, effects of G E were not great between conventional and grazing herds in Canada. The most profound observation was a scaling effect, in which phenotypic differences in yield between cows in conventional herds were approximately 20 percent greater than would have been predicted based on the EBV of their respective sires, while cattle from grazing herds would be expected to differ by about 10 percent less than suggested by sire EBV. This scaling effect seemingly contributed to higher estimates of heritability in conventional herds. Genetic correlations of the same yield trait in different environments were not different from 1, except for fat yield, which at 0.88 was still relatively high. For non-production traits effects of G E were practically non-existent.
A number of factors could explain this lack of observed G E. First, due to the cooler climate and shorter growing seasons in Canada, the herds in this study practiced grazing for only about half of the year. The rest of the time feeding and management would be expected to be similar to the rest of Canada. Therefore, the two systems of management were not as extremely different as could be observed in another country such as the US or between countries. Secondly, all of the herds in both groups were enrolled in DHI and had primarily registered animals. These factors may suggest that other management practices were in common across groups. Another possibility is that expression of genes is not particularly different in quite diverse environments.
The level of G E is too low to justify a separate national sire proving program for graziers, particularly in Canada, where, due to the climate, the number of producers practicing is relatively small. Weigel et al. (1999) concluded that a separate sire proving program for graziers would require sampling of 600 to 700 bulls per year to achieve the same genetic progress as using sires based on a national program for which results were genetically correlated by 0.90. Weigel et al. (1999) also questioned the practice of using sires proven in typically grazing countries such as New Zealand or Ireland, because progeny testing programs in those countries are small and selection intensity would be lost.
Results and Discussion-Production. Means and standard deviations of MEM, MEF, and MEP for control and grazing herds are in Table D.2. As expected, the mean MEM, MEF, and MEP were higher in the control versus the grazing herds and similar to those obtained by Weigel et al. (1999) for herds in Wisconsin. The standard deviations for milk, fat, and protein were 83kg, 3kg, and 1kg higher for the control herds respectively.
A summary of the production data for the regression analyses is in Table D.2 for random subsets and in Table D.3 for quartiles based on herd mean for MEM. For the random subsets for the regression analysis, mean first lactation production and standard deviations were similar to the complete data set. For the quartile analysis the difference between the highest and lowest quartiles was over 4000 kg of MEM for both grazing and control herds.
Coefficients of regression of daughter performance in first lactation on the USDA-DHIA sire PTA for grazing and control herds are in Table D.4. The expectation of the coefficients is 1. For the overall grazing data the coefficients for milk, fat, and protein were significantly less than 1. These results differ from those reported by Weigel et al. (1999) who found only fat for the grazing herds to be significantly different from 1. A possible explanation is that only 5621 records were used in that study, limiting the likelihood of detecting modest differences. The coefficients of regression for the overall data sets are in agreement with the random subsets, in that the coefficients generally do not differ from expectation for the controls, while those for the grazing herds are significantly different from expected values. For the overall data sets the coefficients of regression indicate that actual sire PTA may overstate the true differences between bulls’ daughters in grazing herds. The estimated differences between environments should be conservative for two reasons. First, these records may have contributed to the USDA-DHIA genetic evaluations from which the PTA are derived. Thus a part-whole relationship may have existed and the larger number of control herds could partly account for a higher correlation with the USDA-DHIA PTA. Second, control herds were of similar size to grazing herds (133 cows for controls, 95 cows for grazing), therefore differences could be greater for larger herds that are perhaps more intensively managed.
Coefficients of regression and standard errors for the analyses within quartiles are in Table D.5. For the quartile analysis it appears that milk, fat, and protein yield are adequately predicted by sire PTA for the middle two quartiles for control herds, but are over-predicted for grazing herds, and most severely for the lowest quartile. When depicted graphically, there appears to be little difference in regression coefficients in the middle quartiles for milk, fat, or protein (Figure D.1). Differences between daughters may be over-predicted in the lowest producing control herds for fat and under-predicted in the highest producing control herds for milk, fat, and protein. When regression coefficients of the quartiles are taken together without consideration of the environment, there appears to be a consistent relationship between the regression coefficients and ME milk yield, suggesting at least a moderate scaling effect with increasing milk production. This may partly account for smaller changes in regression coefficients between the inner quartiles, since differences and within-quartile variation might be expected to be less if distribution of records approaches normality.
Estimates of heritabilities and genetic correlations for the four subsets and the weighted average estimates are in Table D.6. Estimates for the same parameters, but for the quartile analyses are in Tables D.7 and D.8. Estimates of the ratio of permanent environment to phenotypic variance ranged from 0.20 to 0.22 for traits in the control herds and 0.18 to 0.21 for traits in the grazing herds. The estimates for control herds were about 2 percent greater than those for grazing herds for each trait. The overall estimates of heritability for milk, fat, and protein were slightly higher for the control herds and also for the upper quartiles for both groups. This is in agreement with previous studies, that have shown less genetic variation in lower producing herds (Cromie et al., 1998, Cienfuegos-Rivas et al., 1999). The heritabilities for all traits for both grazing and controls were similar for the middle quartiles. Across subsets, there was variation among the estimates for each trait, which can probably be attributed to sampling error. Heritiability estimates were more stable across subsets for fat, than either milk or protein.
Overall estimates of the genetic correlation between traits in different environments were significantly different from unity. The estimates obtained were similar to those of Weigel et al. (1999), but due to larger standard errors in that study (approximately 0.1), the estimates were not found to be significantly different from unity. The correlations obtained in this study indicate that there is at least a scaling effect among sires in both environments i.e. the advantage of sires in the control herds is less when their daughters are producing under grazing conditions. However, the correlations are still higher than the value of 0.8 that Robertson (1959) proposed as a value indicative of a biologically important G×E. For the quartile analysis, a correlation of unity existed between the upper and lower quartiles for the control herds. However, the correlations for the grazing quartiles did differ significantly from unity for milk (0.82) and protein (0.85) but not for fat. This indicates that genes for milk, and protein production may be expressed differently in lower producing grazing herds.
Cromie et al. (1988) also found a similar decline in the genetic correlation for milk and protein as the differences in environments became more pronounced. This may not be surprising as cows producing in the lowest quartile grazing herds are probably producing under very extensive conditions. For instance, in the lowest quartile, 86 of the 99 herds were from Louisiana, while there were only 30 of 99 herds in the upper quartile. Due to the very warm summer climate it is likely that pasture shortages occur frequently resulting in less intensive grazing practices. These results are in agreement with the quartile regression analysis, which had relatively low coefficients of regression for milk, fat and protein in the lowest quartile. Genetic correlations for the middle quartiles for both grazing and control for all traits were not different from unity. The high correlations between these quartiles might be expected as the level of management should be similar.
Product-moment and rank correlation coefficients for sires based on their EBV calculated separately for both grazing and control are in Table D.9. The table also includes the rank correlation between the sire’s EBV in both systems and his official November 2000 USDA-DHIA PTA for the same traits. The correlations are based on 792 sires with greater than 5 daughters producing in both environments. Product-moment correlations were similar to the rank correlations for all traits. Product-moment correlations between sires’ EBV in both groups were 0.62, 0.64, and 0.66 for milk, fat, and protein respectively. These are in good agreement with Cromie et al. (1998), who found similar results for sires in high and low input herds in Ireland. The correlations for both grazing and control herds with the USDA-DHIA PTA are higher than the correlations between grazing and control herds, partly because the USDA-DHIA PTA include the information from both grazing and control herds.
For all traits, the rank correlation of the sire’s EBV in control herds and their USDA-DHIA evaluation were approximately 6 percent to 10 percent higher than the correlation between the grazing EBV and the USDA-DHIA evaluation. To quantify the change in rank between sire EBV in both systems an average rank change for milk was calculated. For all 895 sires the average rank change between grazing and confinement was 156. For the top 100 sires based on milk EBV in control herds, the average rank change was 27, and 16 percent of bulls had absolute rank changes of more than 50 (Table 10). For bulls with at least 25 daughters in both environments, Figure 2 has the rank of the bull’s EBV for milk in control herds plotted against the same bull’s EBV in grazing herds. The rank correlation for these bulls was .62. When evaluations were performed within systems, a fair amount of re-ranking occurred. A limitation of this approach is that only the information from within each environment is taken into account when calculating the EBV, leading to differences in accuracy of the EBV, as some sires may have many more daughters in confinement herds than grazing and vice versa. For example, a sire with many daughters in confinement herds may have a high accuracy EBV reflecting his true breeding worth in control environments, but if he has few daughters in grazing, his EBV will be regressed more toward the parent average, which may not accurately reflect his true genetic value as well. Also, these were single-trait evaluations. Multi-trait evaluations may be more accurate, but this was beyond the scope of this project.
Results and Discussion-Health and Reproduction. Means and standard deviations for each of the traits are provided in Table E.2. LSCS was numerically higher for the grazing herds than for the controls herds. Cows on pasture might be expected to have a lower SCS, however there was a disproportionate number of grazing cows to control cows for Southern states, especially Louisiana and Mississippi, where SCS may be higher due to the hot and humid climate (Schutz et al. 1994). Likewise, the reproductive performance of the grazing cows is slightly lower than that of the control cows, but similar to expected benchmarks (Indiana State Dairy Association, 2000). It is possible this could be attributed to the unsuitability of the genetics of these cows for pasture based systems, but more likely can be attributed to a greater majority of grazing cows in Southern states, as well. Nevertheless, 154 days open on average would make seasonal calving very difficult. However, there was significant variation around these means as evidenced by the relatively large standard deviations.
Coefficients of regression of daughter first lactation LSCS on USDA-DHIA sire PTA for SCS for the overall and random subsets are in Table E.3. The expectation of the coefficients was near 1. For the grazing group, only one subset produced a regression coefficient that was significantly less than expectation. One random subset in the control group was significantly greater than 1. The coefficients suggest that daughter differences expected from selecting current sires according to PTA for SCS will be slightly lower in the grazing herds than in control herds. Nevertheless, selection based upon current PTA for SCS should be successful under both environments.
Estimates of heritabilities, genetic correlations, and variance ratios for the permanent environmental effect for SCS are in Table E.4. Heritabilities ranged from 0.1 to 0.15 and are in agreement with previous estimates (Schutz 1994, Banos and Shook, 1990). Differences between heritabilities for the two systems were small and negligible. Since LSCS in grazing and control herds were treated as separate traits, the estimates of genetic correlation indicates the extent to which the two traits are influenced by the same genes, and also the amount of re-ranking expected among sires between the two environments. For the overall data set, the genetic correlation for LSCS of 0.89 was not significantly different from unity, indicating the LSCS is under the control of the same genes in the two environments. Two of the random subsets differed from unity, but this could probably be attributed to sampling error. Permanent environmental effects accounted for about 20 percent of the variation associated with LSCS in both environments. Rank-correlations between grazing and control groups were estimated for 778 sires common to both (Table E.5). The correlation between EBV from grazing and control groups was 0.49. The correlations between the control EBV and the USDA-DHIA PTA were higher (0.7) than the correlation between the grazing EBV and the USDA-DHIA PTA (0.59). The mean rank change between grazing and control herds was 29 for the top 100 sires based on sires’ EBV in control herds, indicating that re-ranking can occur when evaluations are calculated within environment.
Estimates of heritabilities, genetic correlations, and permanent environmental effects for DO, DFS, and SPC are in Table E.6. Heritabilities for all traits were low, ranging from 0.5 – 5%, and are in agreement with several studies (Marti and Funk, 1994, Van Arendonk et al. 1989, Seykora and McDaniel, 1983). Heritabilities were slightly higher for the control herds for all traits, but differences were small. Such low estimates indicate that environmental and non-additive genetic factors contribute most to the variation associated with the traits. For DO, the genetic correlation was different from unity for one subset, but not for the other. The overall estimate, a weighted average of the two subsets by the number of observations, was not different from unity. Sampling error could account for the lower estimate. Estimates of genetic correlation were unity for both DFS and SPC, indicating that the traits are under identical genetic control in both environments. Estimates of permanent environment were around 6 percent for DO, but fluctuated widely for DFS and SPC. This can probably be attributed to the relatively small number of records, and especially repeated records for these traits.
Information on liveweight, body condition scores and weight loss post calving were not available for this study. Aforementioned studies have indicated that as milk production has increased the reproductive capacity of these animals has declined. Failure to match nutritional demands have been suggested as the main reason, with high producing cows simply not consuming enough energy to cope with the demands of production and reproduction simultaneously. This in itself raises many questions for graziers regarding the suitability of high producing cows as the energy intake of grazing cows is less than the intake in confinement. In New Zealand, Kolver et al. (2000) reported that heifers of overseas origin (primarily North American) were unable to maintain acceptable body condition and live weight on an all-pasture diet, compared to New Zealand heifers, questioning the suitability of the overseas Holstein-Friesian genotype for seasonal pasture production systems. In Ireland, Snijders et al. (2001), similarly reported that cows of high genetic merit had a lower body condition score at calving and lost more body condition from calving to first service than medium genetic cows, despite the high merit group having a significantly higher total dry matter intake. If such a similar pattern for body condition and weight loss existed for grazing herds in the US, the ability to re-breed cows rapidly and achieve seasonal calving would be restricted. Mean DO, and DFS reported here, are long and tend to suggest that seasonal calving would be difficult, but the reasons for this are unclear from the information available.
Results of study have been reported in 3 articles for the Journal of Dairy Science and 2 published abstracts from presentations at annual meetings of the American Dairy Science Association. Preliminary SARE reports have led to two articles in the popular press, invitations to lead three workshops on genetics for graziers.In these events, along with the presentation of results from this study, we also discussed crossbreeding of dairy cattle, which is another key component of genetic selection for pasture-based systems. A final presentation of results with advice for producers is under development and will be made available via the Internet.
Banos, G. 1997. INTERBULL routine genetic evaluation for dairy production traits. International Bull Evaluation Service. Uppsala, Sweden.
Banos, G. and G.E. Shook. 1990. Genotype by environment interactions and genetic correlations for somatic cell count and milk yield. J. Dairy Sci. 73:2563.
Boettcher, P. J. and Gibson, J. P. 1997. Estimation of variance of maternal lineage effects among Canadian Holsteins. J. Dairy Sci. 80:2167-2176.
Butler, W.R. and R.D. Smith. 1989. Interrelationships between energy balance and postpartum reproductive function in dairy cattle. J. Dairy Sci. 72:767.
Cienfuegos-Rivas, E.G., P.A. Oltenacu, R.W. Blake, S.J. Schwager, and H. Castillo-Juarez. 1998. Interactions of milk yield between Holstein cows in the US and Mexico. J. Dairy Sci. 81(suppl. 1):80 (Abstr.).
Cromie, A.R., D.L. Kelleher, F.J. Gordon, and M. Rath. 1998. Genotype by environment interaction for milk production traits in Holstein-Friesian dairy cattle in Ireland. Proc. INTERBULL Open Meeting, Jan. 18-19, Rotorua, New Zealand.
Dillon. P, F. Buckley, S. Snijders, and S. Crosse. 1998. The effect of cow genetic index and grass based feeding system on the performance of spring calving Holstein-Friesian cows in second lactation. Proc. Int. Symp. Metabolic Stress in Dairy Cows. Edinburgh, Scotland.
Dimov, G., L.G. Albuquerque, J.F. Keown, L.D. VanVleck, and H.D. Norman. 1995. Variance of interaction effects of sire and herd for yield traits of Holsteins in California, New York, and Pennsylvania with an animal model. J. Dairy Sci. 78:939.
Dimov, G., J.F. Keown, L.D. VanVleck, and H.D. Norman. 1996. Effect of variance of interaction effects of sire and herd on selection for milk and fat yield. J. Dairy Sci. 79:140.
Fulkerson, W. 1997. Are high ABV cows more efficient? Genetics Australia. p. 31.
Galindo, F.A. and Broom, D.M. 1993. The relationships between social behaviour of dairy cows and the occurrence of lameness. J. Brit. Cattle Vet. Assoc. 1:360-364.
Groeneveld, E., and A. Garcia-Cortes. 1998. VCE4.0, a covariance component package for frequentists and Bayesians. Proc. 6th World Congr. Genet. Appl. Livest. Prod. 27:455-456.
Harrison, R.O., S.P. Ford, J.W. Young, A.J. Conley, and A.E. Freeman. 1990. Increased milk production versus reproductive and energy status of high producing dairy cows. J. Dairy Sci. 73:2749.
Henderson, C.R. 1984. Applications of linear models in animal breeding. University of Guleph, Guelph, Ontario, Canada.
Kolver, E. S., A. R. Napper, P. J. A. Copeman, and L. D. Muller. 2000. A comparison of New Zealand and overseas Holstein Friesian heifers. Proc. New Zealand Soc. of An. Prod. 60:265-269.
Marti, C. F., and D. A. Funk. 1994 Relationship between production and days open at different levels of herd production. J. Dairy Sci. 77:1682-1690.
Miesenberger, J., J. Solkner, and A. Essl. 1998. Economic weights for fertility and reproduction traits relative to other traits and effects of including functional traits into a total merit index. Pages 78-84 in Proc. Int. Workshop on Genet. Improvement of Functional Traits in Cattle. Interbull Bulletin No. 18.
Neumaier, A. and E. Groeneveld. 1998. Restricted Maximum Likelihood estimation of covariances in sparse linear models. Genet. Sel. Evol. 30:3-26.
Robertson, A. 1959. The sampling variance of the genetic correlation coefficient. Biometrics 15:469-485.
SAS Institute Inc. 2000. SAS/STAT release 8.01. SAS Institute Inc., Cary, NC.
Schutz, M. M. 1994. Genetic evaluation of somatic cell scores for United States dairy cattle. J. Dairy Sci. 77:2113-2129.
Schutz, M. M., P. M. VanRaden, and G. R. Wiggans. 1994. Genetic variation in lactation means of somatic cell scores for six breeds of dairy cattle. J. Dairy Sci. 77:284-293.
Senatore, E.M., W.R. Butler, and P.A. Oltenacu. 1996. Relationships between energy balance and post-partum ovarian activity and fertility in first-lactation dairy cows. Anim. Sci. 62:17.
Seykora, A. J., and B. T. McDaniel. 1983. Heritabilities and correlations of lactation yields and fertility for Holsteins. J. Dairy Sci. 66:1486-1493.
Short, T. H. and T. J. Lawlor. 1992. Genetic parameters of conformation traits, milk yield, and herd life in Holstein. J. Dairy Sci. 75:1987-1998.
Snijders, S. E. M., P. G. Dillion, K. J. O’ Farrel, M. Disken, A. R. G. Wylie, D. O’ Callaghan, M. Rath, and M. P. Boland. Genetic merit for milk production and reproductive success in dairy cows. Anim. Reprod. Sci. 65:17-31.
Stanton, T.L., L.D. Van Vleck, and M.J.Carabano. 1991. Genotype by environment interaction for Holstein milk yield in Columbia, Mexico, and Puerto Rico. J. Dairy Sci. 74:1700.
Van Arendonk, J. A. M., R. Hovenier, and W. De Boer. 1989. Phenotypic and genetic association between fertility and production in dairy cattle. Livest. Prod. Sci. 21:1-12.
Veerkamp, R.F., G. Simm and J.D. Oldham. 1995. Genotype by environmental interactions: experience from Langhilll. Pages 59-66 in Occasional Publication No. 19. British Society of Animal Science.
Webb, R., P. Garnsworthy, J.G. Gong, R.S. Robinson, and D.C. Wathes. 1998. Consequences for reproductive function of metabolic adaptation to load. Proc. Int. Symp. Metabolic Stress in Dairy Cows. Edinburgh, Scotland.
Weigel, K.A. 1998. Impact of dairy herd management practices on genetic selection decisions. Proceedings National Dairy Genetics Workshop, Mar. 29-31, Orlando, FL.
Weigel, K.A. and A.L. Pohlmann. 1998. Genetic performance of dairy cattle under management intensive grazing conditions. J. Dairy Sci. 81(suppl. 1):8
Genetic correlations <1 for all production, reproduction, conformation and health traits examined, suggest that there is some G×E for production traits between confinement and grazing in the US and Canada. Little evidence was noted for the presence of GxE for fertility or conformation traits. The genetic correlations for production obtained in this study were significantly less than unity, indicating that there is at least a scaling effect. However, based on a genetic correlation of 0.9 between control and grazing herds, Weigel et al. (1999) estimated that graziers would have to sample 500 to 700 young sires each year to achieve the same level of genetic gain available from selecting among the current AI sires sampled annually. Estimates calculated in this study are close to 0.9 for production traits and much higher for fertility traits, therefore a young sire sampling scheme solely to identify sires that excel under grazing, does not appear to be economically feasible given that a high level of genetic gain will be achieved by selecting sires based on current evaluations. However, some graziers may need to be concerned that that a small amount of re-ranking can occur. This is especially true for those graziers with low producing herds. Differences in daughter performance predicted by USDA-DHI sire PTA values may be reduced by as much as half in these herds. Also, a lower correlation for the traits between the upper and lower quartiles for grazing indicates the possibility for re-ranking. Nevertheless, sires will continue to rank nearly the same, no matter the environment. While this study indicates that G×E for grazing and confinement may not be very significant for production traits, other traits such udder health and fertility attain added importance, and certainly have a greater economic value in seasonally calving herds. It is possible that G×E plays a greater role in traits under less additive genetic control as opposed to those with large additive genetic effects for which the accuracy of sire PTA is greater. After many years of selecting solely on milk production traits; milk quality, health, and fertility are becoming more focal aspects of dairying. Premiums for good quality milk, financial losses associated with mastitis, and high cow replacement costs, mean it is important to consider these health and fertility traits in breeding programs. Despite the low estimates of heritability for the traits analyzed, there is still additive genetic variation around the traits. Results of this study suggest that there is little evidence for G×E for LSCS, days open, days to first service, number of services per conception and conformation for the environments defined. Evidence of limited re-ranking of sires EBV for LSCS when evaluations are done within environment do raise the possibility of imperfections of current evaluations for accurate prediction in grazing situations. However, it is likely that selecting among the top sires based on current evaluations will make nearly optimal genetic progress, especially for individual traits in the selection program.
There is some genotype by environment interaction for production traits between confinement and grazing herds in the US. However, a young sire sampling scheme solely to identify sires that excel under grazing does not appear to be economically feasible, given that a high level of genetic gain will be achieved by selecting sires based on current evaluations. Graziers can confidently use bulls tested primarily in conventional dairy environments. Indeed use of artificial insemination sires instead of natural service sires from other grazing herds will result in an increased value of at least $210 per heifer over the course of her life. For a 100 cow grazing herd, this equates to $10,500 per yearly calf crop.
Educational & Outreach Activities
Boettcher, P.J., J. Fatehi, and M.M. Schutz. 2002 (In Press). Genotype-by-Environment Interactions in Conventional versus pasture-based dairies in Canada. J. Dairy Sci. (University of Guelph manuscript).[Appendix C]
Kearney, J.F., M.M. Schutz, P.J. Boettcher, and K.A. Weigel. (Accepted). Genotype by environment interaction in grazing versus confinement dairy production. I. Milk production traits. J. Dairy Sci. [Appendix D]
Kearney, J.F., M.M. Schutz, and P.J. Boettcher. (Accepted). Genotype by environment interaction in grazing versus confinement dairy production. II. Health and reproduction. J. Dairy Sci.[Appendix E]
Boettcher, P.J, J. Fatehi, and M.M. Schutz. 2001. Effects of genotype-by-environment interactions in conventional versus pasture-based dairies. J. Dairy Sci. (Suppl. 1):214. [Appendix A]
Kearney, J.F., M.M. Schutz, K.A. Weigel, P. J. Boettcher. 2001. Evidence for genotype by environment interacton in production traits of US Holsteins under grazing versus confinement. J. Dairy Sci. (Suppl. 1):214. [Appendix B]
Schutz, M.M. Genetics and Crossbreeding for Graziers. Midwest Grazing Conference. Dubuque, IA. July 2001.
Schutz, M.M. Genetics and Crossbreeding for Graziers. Southeastern Wisconsin Grazing Conference. Kaukauna, WI. April 2002. [Copy of slide presentation in Appendix F]
Popular Press articles resulting from interviews
There’s profit in grass. Northeast Dairy Business. September 2001.
Genetics of confinement versus pasture-based systems. Holstein Journal 64(7):40-41, 2001.
Areas needing additional study
During this study a number of issues came to light that need further study. The key issue is that in order to develop genetic selection programs and selection indexes that are appropriate for graziers, there must be an estimate of the economic importance of specific traits relative to grazing versus confinement. Release of genetic evaluations for fertility traits by USDA in February, 2003, will provide a sorely needed tool for graziers practicing seasonal calving. However, it remains to be determined what level of gain can be achieved for such a lowly heritable trait. Also, work must be undertaken to determine the relative value of different sources of genetic material, such as breeds or strains from around the world that will work in different dairy production environments. We are continuing work to determine relative economic weights of specific traits in grazing environments.