Goats play an important role in the economy of traditional pastoral production systems in northern Kenya. Goats are kept for both meat and milk production, but overall contribution of goats milk to total milk consumption in pastoral households is limited due to their low milk yields which usually range between 30 and 100 litres per lactation. Hence, the primary output from pastoral goat flocks is meat production both for subsistence consumption and for sale or barter. According to a survey conducted by Schwartz (1986) in Marsabit District, northern Kenya, goat meat can account for more than 70 percent of all meat consumed, and sales of goats and of goat skins can contribute up to about 40 percent of the total annual cash income of pastoral households.
Like other components of performance that determine overall biological productivity of goat herds, such as reproduction and survival, growth and liveweight development in all animal categories are greatly affected by the seasonality in pasture forage production that is typical of semi-arid rangelands in East Africa. The scope for alleviating restrictions imposed on animal performance by the seasonally and annually varying supply of nutrients is very limited. Since pastoral producers lack the ability to manipulate the primary productivity of the grazing system, imbalances between feed supply and nutrient demand from the herd have traditionally been overcome by adaptive management, such as the movement of animals onto pastures with adequate forage production and flexibility in the timing of herd offtake. However, the distinctive advantage derived from herd mobility for compensating nutrient shortages is lost in the transition to sedentariness, an ongoing process which has repeatedly been observed among pastoral communities in semi-arid regions of Kenya (Grandin, 1988, 1991; Fratkin, 1992; Schwartz et al., 1995; Roth, 1996). The reduced mobility of pastoral herds introduces an additional constraint on animal productivity, with repercussions on the welfare of pastoral households.
In this situation, restricting breeding to a short period in a year could help to synchronize seasonal nutrient supply from pastures with the most physiologically demanding stages of the production cycle (Delgadillo & Malpaux, 1996; Walkden-Brown & Restall, 1996). Peak nutrient demand generally occurs from late gestation until weaning, and nutritional deficits during this stage can be expected to lead to low birth weights and milk yields, slow growth, and an increased risk of death of suckling kids. Although local breeds such as the Small East African (SEA) goat have the potential to breed non-seasonally, breeding does may exhibit periods of anovulation or anoestrus as a result of poor feeding conditions. This may be exacerbated if liveweight recovery of dams is inhibited by nutritional stress during the postpartum period, and may cause losses in productivity in subsequent reproductive cycles through lowered fertility, prolificacy, milk production, and survival.
The objective of this study is to investigate whether liveweights and growth performance of youngstock in a herd of SEA goats maintained under pastoral management can be improved by introducing a controlled seasonal breeding regime. Given the importance of liveweight recovery and development of dams for their subsequent reproductive and milk performance, an additional goal is to examine whether, over a reproductive cycle of one year duration, liveweight development of breeding females would benefit from such a change in production strategy.
Data for this study came from a herd of SEA goats maintained at the Ngare Ndare Research Station of the University of Nairobi in Isiolo District, northern Kenya, over a four-year period from 1984 to 1988. [page 64↓]The study area is located 25 km west of the district capital Isiolo at an altitude between 1000 and 1300m. The climate is semi-arid with unreliable rainfall (long-term annual average at Isiolo township: 615 mm) distributed over two wet seasons, the long rainy season extending from March to May and the short rainy season from October to November. The predominant pasture types are annual grassland, dwarf shrub land and Acacia sp. dominated thornbush of medium density. Limited areas of riverine woodlands and perennial grassland on a floodplain serve as dry season grazing reserve.
Details of the experiment with regard to herd management and data recording were described before in Chapters 2 and 3. Briefly, breeding groups of 18 does each, with a balanced age and weight structure, were established. The experiment was initiated at the end of January 1984, when a buck was introduced into the first breeding group for two months. Afterwards, the buck was transferred to consecutive groups for the same duration, so as to achieve year-round mating, kidding, and weaning. Kids were weaned at an age of 16 weeks. For the purpose of analysis, data from the individual breeding groups obtained over the 40 months duration of the experiment were grouped into six different mating seasons. Three complete production cycles, ranging from mating until the time at which youngstock had reached an age of two years were obtained for five of these six consecutive two month mating periods. The sixth period had only two complete cycles, because the last breeding group of the experiment, which was set up by the end of March in 1987, had to be discarded due to incomplete records. The experimental treatment thus consisted of six mating seasons, the first one, labelled as mating season 1, ranged from February to March, and the last one, labelled as mating season 6, ranged from December to January. Mating seasons 4 and 5 (August to October, and October to December, respectively) had to be assumed to have taken place over a period of three months due to a delay of one month which occurred in setting up the first breeding group of mating season 4 in 1984.
Time in the experiment was divided into periods of two weeks, each period starting and ending on a Monday. Routine measurements of weight (kg) of all animals, as well as milk yields of dams were carried out at the beginning of each two-week period. Adult animals were weighed using a weighing crate and measurements were made to the nearest 0.5 kg. Kids were weighed immediately after birth using a spring balance to the nearest 0.25 kg. A total of 8547 measurements were obtained on liveweight and milk production of does; 9837 observations were available on liveweight development of youngstock.
Pasture condition was judged every two weeks using a subjective phenological pasture condition score ranging from 1 to 4 based on greenness and abundance of the herblayer (range condition score [I]), including grasses, herbs, and small dwarf shrubs. Scores of 1 to 2 generally occurred during the dry seasons, and 3 and 4 during and immediately after the rainy seasons. The condition score for the herblayer was upgraded to a maximum score value of 5 to integrate the contribution of bushes and trees with regard to browse availability and the production of high quality litter such as leaves, flowers and fruits (range condition score [II]).
Growth rates of kids, expressed as average relative daily liveweight gains (g·kg-0.75·day-1) were analysed from birth until two years of age at 4 weeks intervals. Relative live weight gain was calculated as the difference between two consecutive live weight measurements divided by the length of each time interval (28 days) and the average of the two live weight measurements raised to the power of 0.75. Animals with complete records over the follow-up period of 104 weeks thus had 27 observations. In order to reduce both execution times of model estimations and the complexity of growth curves fitted to the data, separate analyses were carried for the time series from birth until 50 weeks, and from 52 until 104 weeks of age.
Average absolute daily live weight gains (g·day-1) over time intervals of 16 weeks each were studied from birth until 96 weeks of age (resulting in 6 observation times for subjects with complete records). These were computed as the difference between two consecutive live weight measurements, 16 weeks apart, divided by the length of the time interval (128 days). Liveweight development of kids (kg) was studied at 8 week intervals, ranging from birth until two years of age. Hence, subjects with complete records had 14 observation times. For all the foregoing traits, observations for immature females which were first mated before reaching the age of 104 weeks were considered to be censored observations from the time at mating onwards, and the corresponding records were deleted from the data set.
Body weight (kg) development of does during pregnancy was studied at biweekly intervals over a period of 52 weeks, such that subjects with complete records had a maximum of 27 observation time points in the analysis. Only those animals with live-birth events were considered. A time series was constructed from the body weight measurements of each doe, ranging from the recording date at the onset of mating to the recording date 52 weeks later. Whenever does were rebred or were found to be pregnant before the end of the current reproductive cycle, observations from this time point onwards were considered to be censored and deleted from the data set. In order to study relative live weight changes over the reproductive cycle, the same [page 65↓]series of measurements was used to compute percentage change in live weight at each time point, relative to body weight observed at mating.
In studying patterns of growth and liveweight development, liveweight is a variable measured repeatedly over time in the experimental units. Principally, this type of sequential, correlated observations can be analysed with multivariate regression techniques. However, longitudinal studies typically give rise to special data structures which do not meet model assumptions imposed by such methods (Verbeke, 1997). A very often encountered problem in the analysis of growth data is that individual animals may be observed a different number of times because they were withdrawn from the experiment for some reason or died prior to the final measurement time. Moreover, time points at which repeated measurements are taken may not be the same for all subjects on study, and the intervals between observations may be different as well. An alternative to multivariate regression which is able to accommodate incomplete and unequally spaced observations is the linear mixed-effects model (Laird & Ware, 1982; Cnaan et al., 1997). This approach is also well-suited for modelling covariances among observations on the same subject and for handling heterogeneous variances over time (Wolfinger, 1996).
The general linear mixed model framework, as implemented in the SAS procedure MIXED (SAS Release 6.12, 1996), was used to fit growth curve models to longitudinal growth and liveweight development data. The aim was to characterize patterns of growth and liveweight change over time for subjects in the six different mating season groups, and to investigate the effects of other covariates on these patterns. The basic premise in growth curve models is that there is a functional relationship between the observed response, a treatment effect, and time. One possible way of approximating the actual functional form of the relationship is through a natural polynomial or polynomial splines in time (Rowell & Walters, 1976; Smith, 1979; Cullis & McGilchrist, 1990), and this approach was adopted in the present study.
Two sources of variation that are potentially present in longitudinal data can be distinguished. First, there is subject to subject variation in the mean response value which can be taken into account by including a random subject factor into the model. It is this between-subject random effect which causes the correlation within each subject’s observations. The second random component is the within subject error. This random effect consists of measurement errors and the actual variability of the individual subject’s response (Jones, 1993). Typically, whenever a response variable is measured repeatedly over time, there is the possibility of serial correlation. Serial correlation is present when measurements within a subject taken close in time are more highly correlated than measurements taken far apart in time. However, Diggle et al. (1995) point out that, although serial correlation would appear to be a natural feature in any longitudinal data model, in many applications its effects may be dominated by the combination of random subject effects and measurement errors. A general description of the models used for analysing longitudinal data is given below.
Measurements (absolute and relative growth rates and liveweight changes calculated from liveweight measurements, as well as liveweight measurements themselves) were obtained from the same unit of observation (individual kids or does), i , i=1,..., n , in the jth treatment group, j=1,..., q, at p pre-specified time points, and subjects were assigned randomly to one of the q treatment groups. Then, following Laird and Ware (1982), the general linear mixed model for a sequence of p measurements on subject i is
y i = a p ix1 vector of the response variable for subject i
X i = a known p ixq design matrix
β = an unknown qx1 vector of fixed effects coefficients
Z i = a known p ixr design matrix
γi = an unknown rx1 vector of random subject coefficients, assumed to be independently distributed across subjects with distribution γi ~Ν(0, σ²B), where B is the between subjects covariance matrix
εi = a p ix1 vector of within subjects errors assumed to be distributed as εi~Ν(0, σ²W i), where W i is the within subjects covariance matrix
The columns of the model matrices X
i and Z i consist of indicator variables corresponding to the levels of the fixed and random subject effects, respectively. The between individuals design matrix X i can specify polynomial powers of the times of observation, treatment and/or classification factor effects and their [page 66↓]interactions with polynomial terms, as well as time-changing covariates such as the range condition scores in the present study (Jones, 1993; Cnaan et al., 1996).
The general linear mixed model (10) extends the general linear model by allowing a more general specification of the covariance matrix of εi. It allows for both correlation and non-identicalness, although normality is still assumed. The second difference lies in the addition of the known design matrix, Z i, and the vector of unknown random subject effects, γi. These between subject components of variance can be used to model the random deviations of the intercept and possibly higher degree polynomial trends in time for subject i from the subject's group mean intercept and higher degree polynomial powers of the times of observation. The linear mixed model formulation is very general since different subjects can have different numbers of observations as well as different numbers of observation times. Therefore, subjects with incomplete data do not need to be discarded from the analysis (Cnaan et al., 1996). Equation (1) can further be expanded to incorporate additional random effects by defining a p ixv design matrix, U i, of random effects and a corresponding unknown vx1 vector, τi, of random effects coefficient s i. The random effects, τ i, are assumed to be identically and independently distributed as τi~N(0, σ²I), where I denotes the pxp identity matrix. In the present study, the effect of production cycle (three levels) and its interaction with the mating season treatment were considered as such random effects and included in all initial models fitted to the data.
Model construction and data analysis encompassed several steps. As a starting point, for each response variable individual profiles and group mean curves were plotted to obtain a visual picture of possible models for the data, including an appropriate variance-covariance structure for the within-subject errors. Based on this preliminary inspection, various models that included random subject effects and serial correlation were computed. Since serial correlation can partially be confounded with the random between subject effects, the procedure recommended by Jones (1993) was used to find an adequate model. It involves fitting four different models, one with uncorrelated (compound symmetry) error structure, one with random subject effects alone, one with a serially correlated within subject error structure alone, and one with both random subject effects and serially correlated error structure. Two additional models, one with and one without random subject effects were fitted to the data using the most general form of heterogeneous variances and covariances, that is, an unstructured covariance matrix. Akaike’s Information Criterion (AIC) was then used to select the best fitting model. The serial correlation structure used was a heterogeneous first-order autoregressive structure (ARH in the SAS PROC MIXED terminology). PROC MIXED uses the parameterisation illustrated by the following 3x3 matrix:
Here, σ2 1, σ2 2, and σ2 3 are the variance parameters at each of three consecutive measurement times, and ρ is the autoregressive parameter. Thus, with the ARH covariance structure a different variance is estimated for each measurement time, such that variance heterogeneity over time can be accounted for in the analysis.
Table 4.1 summarizes the predictor variables included in preliminary analyses of longitudinal data. Selection of suitable final models was based upon fitting successive nested model versions, and omitting variables from the linear predictor when the likelihood ratio (LR) statistic indicated that the model including an additional parameter did not fit significantly better than the simpler model. Fixed-effects terms and interactions among fixed-effects were excluded from the model when the probability of obtaining a greater χ²-value than the observed LR test statistic from a χ² distribution with r degrees of freedom was greater than 0.15, r denoting the additional parameters that have to be estimated in the larger of the two models. The same procedure was applied to test for the significance of random effects, except that the p-value for the LR statistic was obtained by taking half of the probability of a greater χ² from a χ² distribution with one degree of freedom, given that only a one-sided test is needed in this case and only one additional parameter is estimated upon including a random effect into the model.
When fitting polynomial growth curves, multicollinearity between different polynomial degrees of a quantitative predictor variable can occur and cause computational as well as interpretational difficulties (Steel and Torrie, 1980). Since the polynomial representation of response over time used in the present context resulted in high pairwise correlations between powers of time, a matrix of orthogonal polynomials in time was constructed following the procedure described in Draper and Smith (1981) and used in fitting response curves. In order to control the experimentwise error rate at the prespecified level of α=0.05 when making multiple comparisons of factor level means, both the Bonferroni and Tukey multiple comparison procedures were used. The procedure giving the narrower confidence limits was then chosen to report [page 67↓]significance probabilities of differences in factor level means. This choice is proper since it does not depend on the observed data (Neter et al., 1996). Tests of hypotheses concerning specific linear combinations of parameter estimates were obtained by specifying appropriate estimate and contrast statements in the MIXED procedure.
Daily weight gains
The analysis of growth rate in kids in terms of relative (g·kg— 0.75·day— 1) and absolute (g·day— 1) daily weight gains revealed neither a significant component of variance due to dam [mating seasonxproduction cycle], nor significant individual differences (random kid effect). Also, both maternal parity stage and postpartum weight had no influence on response patterns. Interestingly, differences in relative body weight gains between female and male kids were very small and insignificant in all the analyses performed on this trait. In contrast, absolute daily growth rates differed markedly between male and female offspring.
|Figure 4.1. Estimated relationships (least-squares means) between daily weight gains by mating season (MS) and age of kids from a) birth until one year of age, and b) from one to two years of age.|
Relative daily weight gain data were modelled as a polynomial function in time differing according to a) mating season, and b) litter size, milk yield until weaning, and lagged range condition scores [I] and [II] (Tables 4.2 and 4.3). The decision to fit separate models in assessing the latter effects was motivated by the fact that they are nonmanipulable or observational classification factors, and that they had been found to be influenced by, or confounded with, the control variable determining functional relationships in this study, i.e. mating season. Models developed to assess the effects of litter size, milk yield until weaning, and lagged range condition scores [I] and [II] were restricted to the period from birth until the yearling stage. This was done primarily because the assessment of range condition was discontinued in November 1987 before surviving kids born in early 1986 and thereafter had reached an age of two years. Additionally, litter size and milk yield until weaning did not significantly affect relative daily weight gain beyond one year of age.
Estimates of absolute growth rate were obtained by fitting an additional model to average body weight gains from birth to 96 weeks for a total of six consecutive life stages, each of 16 weeks duration. Here, only mating season and sex of kid systematically affected observed outcomes (Table 4.4).
Estimated relationships between kid age and relative daily weight gains by mating season are depicted in Figure 4.1. A fifth order polynomial in time was necessary in order to adequately represent relative weight [page 71↓]gains until one and two years of age (Table 4.2). The highly significant interactions of mating season with the various polynomial time trends resulted in completely different response curves for each for each of the six mating season groups. A notable feature of the analysis from birth to one year of age is that no serial correlation between adjacent observations on the same subject was present in the data. The covariance structure used was therefore an unstructured matrix, with zero covariance among subsequent measurements, but heterogeneous variances across time. Although very small in magnitude (ρ=-0.12), correlated within-subject errors occurred from year one onwards, and a heterogeneous first-order autoregressive structure was found to be the most appropriate structure for modelling relative weight gains up to two years of age. Thus, during the period from birth until 52 weeks, intensities of growth were serially independent across age intervals and appeared to be determined mainly by environmental conditions, as is illustrated by the large and recurrent fluctuations in responses over time (Figure 4.1). Beyond 52 weeks of age, the small negative first-order autoregressive coefficient lead to an alternating pattern of autocorrelations, reflecting a tendency for responses to oscillate about a slightly positive mean relative growth. However, the correlation between observations separated by two time steps is almost nil ((-0.12)²= 0.014), indicating a weak dependence of current on past growth performance. The covariance parameter estimates for both models in Table 4.2 clearly illustrate the variance heterogeneity in relative liveweight gain over time, and underline the importance of using a heterogeneous covariance matrix in modelling within-subject errors. Finally, there was a small, but highly significant variation attributable to the effect of production cycle, and to its interaction with mating season (p <0.01).
Generally, it should be emphasized that the number of animals for which there were performance records diminished considerably in age intervals beyond 70 weeks of age, either because animals were taken off for slaughter or, in the case of replacement females, because they were transferred into the breeding herd. Daily weight gain estimates over this time period were therefore less reliable than those for age intervals between birth and one year of age. This is particularly true for estimates relating to mating season 6, for which no data had been obtained during the first production cycle.
|Figure 4.2. Relationships (least-squares means) between daily weight gains and age of kids, according to a) milk yield until weaning and b) litter size (estimates were obtained from model a) in Table 4.3).|
Kids born in mating season group 5 displayed the highest initial propensity to growth (0 to 4 weeks), with an estimated growth rate of 53 g·kg— 0.75·day— 1, closely followed by kids in groups 4 and 6 (both 52g·kg— 0.75·day— 1). In these three groups, births took place between the months of December and May, during which, in general, favourable pasture conditions prevailed. In contrast, kids born during the long dry [page 72↓]season (group 2) grew at a significantly lower rate of 40g·kg— 0.75·day— 1 (p <0.05) during the first four weeks of life. Intermediate weight gain levels of 48 and 43g·kg— 0.75·day— 1 during the same age interval were estimated for mating season groups 1 and 3, respectively.
The fastest decline in growth rate after 4 weeks of age occurred in group 1, in response to declining milk yields as the long dry season progressed. Within ten weeks from birth, growth slowed down to a rate of 5 g·kg— 0.75·day— 1, which corresponded to only 41 percent of the mean weight gain achieved by kids in all other groups in the age interval of 8 to 10 weeks (p <0.01). As expected, the rate of decline in relative growth was less pronounced when kids were born during the short rains (group 3) and just prior to (group 4) or during the long rainy season (group 5). In the latter treatment groups, nutritional constraints did not seem to significantly lower early growth in kids, since daily weight gain in kids remained well above 5g·kg— 0.75·day— 1 until at least 24 weeks of age. Despite their low initial intensity of growth, offspring born in mating season group 2 displayed comparatively high growth rates of about 13 (8-10 weeks) to 10 g·kg— 0.75·day— 1 (20-22 weeks) after the sixth week. In this group, kids probably benefited both from increased levels of available nutrients as forage growth and quality improved at the start of the short rainy season, and from concurrent increases in milk yields of does.
Kids in mating season group 1 showed a remarkable recuperative capacity around the time of weaning, when quantity and quality of forage on offer began to improve at the onset of the short rains. Relative growth rates more than doubled between 14 and 20 weeks of age and reached a maximum value of 12 g·kg— 0.75·day— 1 at 24 to 26 weeks. Under the same environmental conditions, compensatory gains achieved in group 6, after the retardation in growth experienced during the long dry season between 16 and 28 weeks of age, were much less pronounced and levelled-off at approximately 5 g·kg— 0.75·day— 1 . With a displacement in time of two months, the estimated relative weight gain curve for mating season group 5 followed a similar pattern. Except for the peaks in growth rate at about one and two years of age (8 and 12 g·kg— 0.75·day— 1, respectively) observed in group 4, the response profiles in Figure 4.1 generally indicate that the rate and persistency of compensatory gains after periods of nutritional stress tended to decline with advancing age. However, even at late stages (>52 weeks) were kids observed to be able to more than offset delays in growth as well as weight losses by considerably increasing growth intensity during the rainy seasons, where maximum relative liveweight gains of about 5 g·kg— 0.75·day— 1 were attained. Responses in mating season group 6 differed noticeably from this overall pattern, in that liveweight development slowed down rapidly beyond one year of age, and the capacity to fully compensate past poor growth performance was virtually absent. In this group, nutritional constraints experienced at early developmental stages (16 to 30 weeks) seemed to have caused permanent stunting from about 64 weeks onwards.
Models fitted to evaluate growth performance in terms of litter size, milk yield until weaning, and range condition scores are presented in Table 4.3. Trends across time in factor level means for all variables considered differed noticeably, as may be seen from the significant interactions of litter size, milk yield until weaning, and range condition scores with polynomial time terms and from the estimated response profiles displayed in Figures 4.2 and 4.3. Likewise to the model fitted to relative weight gains until 52 weeks that contained mating season as the sole explanatory variable, no serial correlation between adjacent measurements on the same subject could be detected, but again, variances were found to be heterogeneous across time. Variation among production cycles was small, but highly significant (p <0.01), and therefore was included in the final models.
Clear superiority in growth until 10 weeks of age was shown by kids which had least restrictions in terms of milk availability (≥52.5 kg), with relative weight gains that declined from initially 54 g·kg— 0.75·day— 1 to 14 g·kg— 0.75·day— 1 at 8 to 10 weeks of age (Figure 4.2). Growth intensity was much lower (p <0.05) in kids born to does producing less than 22.5 kg milk until weaning, with estimated rates of gain of 40 g·kg— 0.75·day— 1 within the first two weeks of life, and 10g·kg— 0.75·day— 1 in the 8 to 10 weeks age interval. Until 10 weeks of age, a significant linear trend (p <0.01) in relative growth rates across milk yield levels was observed.
An interesting feature emerged as kids approached the weaning stage. Kids which had access to very low quantities of milk during the suckling period (<22.5 kg) showed an increasing intensity of growth from 14 weeks onwards, reaching a maximum of 10 g·kg— 0.75·day— 1 at 20 to 22 weeks of age, well above that achieved in all other milk yield classes in which growth slowed down steadily with advancing age. Offspring in the former milk yield class were able to maintain the highest growth intensity until at least 34 weeks. As is evident from the response profiles depicted in Figure 4.2, the linear trend in relative weight gains across milk yield levels disappeared beyond about 10 weeks of age, which indicates that body weight development in kids became increasingly independent of the amount of milk produced by their dams. This is particularly true for kids which received very low milk quantities, such that they probably had been effectively weaned at an earlier stage.
Single born kids grew more intensively than twin born kids before 10 weeks of age, although this difference was statistically significant (p <0.05) only within the first four weeks of life. From 20 weeks onwards, this trend was reversed and relative growth rates of twins exceeded those of singles by 13 (20-22 weeks; p <0.05) to as much as 68 percent (36-38 weeks; p <0.05). By 42 weeks of age, differences in rates of weight gain due to litter size were no longer significant.
|Figure 4.3. Effect of lagged median range condition scores on daily weight gains of kids until 50 weeks of age.|
Pasture condition, as measured by the two range condition scores, had a considerable effect on relative growth performance of kids (Figure 4.3). Differential effects of RC score [I] levels were generally less pronounced than those of RC score [II]. Until 14 (RC [I]) and 10 weeks of age (RC [II]), relative weight gains followed the expected pattern and were found to be an increasing function of range condition in each age interval. During the suckling period, this relationship was certainly mediated through the effect of pasture quality on milk production. For lagged RC score [I], significant differences were found between levels 1 and 4 at all ages, between levels 1 and 3 from 20 to 38 weeks, and between levels 2 and 3 from 12 to 38 weeks (p <0.05). Response patterns for levels 1 and 4 became inconsistent beyond 38 weeks of age. This was mainly due to the influence exerted by extreme values in weight gains, observed in both of these levels at the end of the time frame, on the estimation of the respective response curves. These would have been smoothed out upon including later age intervals into the analysis.
Several inconsistencies occurred with respect to estimated effects of lagged RC [II] scores on relative weight gains beyond 10 weeks of age. Firstly, estimates for score level 2 in each age interval considered were based on few observations (N <10) and, therefore, were probably seriously biased. This might account for the fact that in this case relative growth rates were predicted to increase, after 14 weeks of age, above those achieved at score values of 1, 3, and 5. Secondly, the sharp decline in response for level 5 from 8 weeks onwards was unexpected, and no clear-cut explanation for this pattern could be identified. The same was true for the estimated trend in weight gains for the third level after 30 weeks of age. Nevertheless, until 42 weeks, relative growth was significantly lower under very poor pasture conditions (level 1) than under favourable conditions (level 4, p <0.01). From 8 to 44 weeks, lagged RC [II] score values of 3 and 5 also led to slower weight gains than score values of 4 (p <0.05).
A summary of the model fitted to evaluate absolute daily weight gains in terms of mating season and sex of kid is given Table 4.4. The estimated covariance parameters show that average daily weight gains achieved in consecutive age intervals were only weakly correlated (ρ=-0.21). Two important features emerged from the Table 4.4. Tests of fixed effects and estimates of variance components and covariance parameters for the model fitted to daily weight gains of kids (g×day-1), accounting for the effects of mating season and sex.
analysis of absolute weight gains in comparison to the results reported previously for relative growth rates in kids. Firstly, the fact that sex of kid significantly affected average daily weight gains in absolute, but not in relative terms, shows that the faster growth rate observed in males is probably entirely due to their higher initial liveweight state, and not to a higher intrinsic propensity for growth. Secondly, a much larger part of the variation observed in responses was attributable to the effect of production cycle, and to its interaction with the mating season treatment effect, than was the case for relative weight gains. The latter variance components cause a vertical shift in BLU-predicted response profiles (i.e., in the intercepts), but do not affect the shape of factor level trends over time. And if it is true that, ceteris paribus, absolute growth rates are primarily a function of liveweight state, then these variations could, to a large extent, be explained by differences in birth weight between replications of the experiment.
Table 4.5 provides estimates of average weight gains by mating season group and sex. Average preweaning gains of about 83 g·day— 1 were achieved when suckling took place between end of October and end of June (groups 3, 4, and 5), and diminished considerably with further displacement of the weaning stage into the long dry season (groups 6 and 1). Compared to offspring in the former groups, suckling kids in group 1 gained weight at a significantly lower rate of 59 g·day— 1, but were able to compensate for their poor preweaning performance by maintaining growth at similar levels during the following eight months of life. Almost identical average daily postweaning gains were achieved by kids born in mating season group 2. Over the same development stage, decreasing pasture quality in the long dry season caused a significant retardation in growth in groups 5 and 6. However, net average weight losses only occurred during the second year of life in mating season group 2 (48-64 and 80-96 weeks), 4 (80-96 weeks), and 6 (64-80 weeks). In all three cases, these weight losses were associated with nutritional stress experienced during the dry seasons.
Overall average daily weight gains from birth until 48 weeks and from 48 to 96 weeks did not differ significantly across mating season groups. Similar performances until 48 weeks of age were observed in groups 1 to 3, and in groups 4 to 6. On average, rates of weight gain were about 11 percent lower in the latter three groups (46 versus 56 g·day— 1). Beyond one year of age, growth rates reduced to values between 20 and 30 g·day— 1. Average daily gains of male kids were significantly higher than those of female kids until about 32 weeks of age. Slightly higher gains in females than males were observed between 48 and 64 weeks, though the difference was not significant. Thereafter, the trend was reversed and, again, males grew faster than females.
Body weight of kids
Statistics that summarize important features of the analysis of kid body weights until two years of age are given in Table 4.6. The dam and kid variance components were found to be nonsignificant. The influences of milk yield until weaning and litter size on weight development of kids were assessed in a separate model, since these effects were confounded with those of mating season and parity. The production cycle variance component and its interaction with the mating season treatment effect had a highly significant influence on model predictions. In both models fitted these sources of variation were small. However, it should be emphasized that they represent deviations of the intercepts of (production cycle specific) response profiles from the population mean intercept, i.e. the population mean birth weight, which itself is small in magnitude and has a smaller variance compared to body weight estimates at later ages, as is apparent from age-specific variance estimates given in Table 4.6.
|Figure 4.4. Estimated growth curves (least squares means) of kids by mating season group (MS).|
Estimated growth curves by mating season, milk yield until weaning, and litter size are depicted in Figures 4.4 and 4.5. Corresponding least squares means and their associated standard errors at selected ages are given in Table 4.7. There was no significant effect of mating season on birth weights. However, does joined during the long dry season in August/September (group 4) gave birth to the lightest kids, whereas kids born to does that were joined at the beginning or end of the short rains (groups 5 and 6) were about 12 percent heavier. Being weaned at the end of the long dry season, kids in the former group were also more than 30 percent lighter at 16 weeks of age than those weaned at the end of the long rains (group 5). As was already indicated by the differences in relative growth rates reported above, maximum weaning weights were generally achieved when suckling took place between end of October and end of June (groups 3 to 5). However, the response profiles in Figure 4.4 show that any disadvantages of low birth and weaning weights incurred by kids in the first two mating season groups were more than offset during a phase of intensive growth observed between weaning and one year of age.
Body weight estimates in Table 4.7 indicate that by 56 weeks, relative differences across mating season groups had considerably diminished. At this age, the largest difference observed was between animals in group 2 and those in group 6, the former being heavier by 2.8 kg or 13 percent. This trend persisted at later ages and, consequently, body weights did not differ significantly between groups 1 to 5 until 88 weeks. In contrast, offspring in mating season 6 never recovered from the nutritional constraints experienced shortly after weaning and, beyond one year of age, had considerably lower liveweights than animals in all other treatment groups. At two years of age, the discrepancy of body weight in group 6 animals compared to others ranged from 3.1 kg (13%, group 4) to a as much as 7.8 kg (33%, group 2).
First parity young weighed significantly less at birth than second parity young, and continued to do so up to 24 weeks of age (p <0.05). Polynomial contrasts revealed a significant cubic trend at birth (p <0.05), and significant linear and quadratic trends (p <0.01) at all later ages in body weights with increasing parity of dam. Kids born to does with at least four prior kiddings grew markedly slower and were lighter from 8 weeks onwards than lower parity kids. Fourth and greater parity young were 4, 14, and 19 percent lighter at birth, one and two years of age, respectively, than second born young which displayed the best overall growthperformance.
Sexual dimorphism in body weights was considerable and became more pronounced with advancing age. The male to female weight ratio increased from 1.05 at birth to 1.13 and 1.17 at one and two years of age. Single born kids were found to be significantly heavier at all ages than twin born kids. The discrepancy in body weight in twin kids initially increased from 7 percent at birth to 22 percent at weaning, but then, due to a faster postweaning growth rate in single kids, reduced to 14 percent by one year of age. Referring to the [page 77↓]estimated body weights at two years of age, litter size had a substantial effect on growth performance; surviving singles being heavier than twins by almost 4 kg. Despite their superior postweaning growth performance, twin born did not fully compensate for their lower initial birth weight and growth intensity.
The amount of milk available to kids during the suckling period had a marked effect on their body weight development up to about one year of age. Thereafter, patterns became inconsistent and, by 96 weeks of age, differences due to milk yield of dam until weaning failed to reach statistical significance. Does with poor milk yields until weaning (<22.5 kg) gave birth to the lightest kids, whereas kids born to does producing more than 52.5 kg milk were approximately 16 percent heavier than the latter. Polynomial contrasts revealed a significant (p <0.01) linear trend in birth weights with increasing milk yield level. At later ages up to 32 weeks, a significant (p <0.01) quadratic trend component was also detected. From the body weight estimates in Table 4.7, it is evident that after 56 weeks surviving kids in the lowest milk yield class had almost fully compensated for their poor preweaning growth performance. By two years of age, body weight differences across milk yield levels did not exhibit any systematic pattern.
Statistics summarizing important features of the analysis of liveweight development of does over a reproductive cycle of one year duration are presented in Table 4.8. This response was modelled as a polynomial spline function in time differing according to mating season and parity. The spline function consisted of a fourth degree base function, and a cubic segment joining at synchronized week 20, coinciding with the beginning of the recording period immediately preceding parturition. The inclusion of the linear trend component (t-20)+ allowed the introduction of a discontinuity into the response curves and thus to [page 79↓]accommodate for the characteristic decline in body weights observable immediately after parturition. A significant part of the variation in liveweights was due to the random effect of production cycle and to its interaction with the mating season treatment variable (p <0.01). In contrast, the subject variance component was found a nonsignificant source of variation (p>0.20).
Results of the model fitted to the percent change in individual liveweights at each time point relative to that measured 22 weeks prepartum (corresponding to the approximate time at conception) are given in Table 4.9. As before, mating season and parity stage were found to have a systematic effect on outcomes. Also, theeffect of mating season on relative weight changes varied substantially across production cycles, as is apparent from the comparatively large (and highly significant, p <0.001), mating seasonxproduction cycle component of variance. As it was set to zero by the REML estimation procedure, the random production cycle effect was dropped from the final model.
The effects of lagged median range condition scores [I] and [II] at each time point were statistically significant, but the respective estimated response profiles did not lend themselves to a meaningful interpretation,
and therefore both effects were omitted from consideration. Since type of birth was confounded with mating season and dam parity stage, its effect on liveweight development of fertile does was studied separately. The approach followed was to include parity and its interaction with litter size as explanatory factors into the model, and then to obtain simple effects response curves for does with single and twin birth events within each parity class. Liveweight development was again assessed both in absolute and in relative terms, and statistical summaries of this part of the analysis are presented in Table 4.10.
Plots of predicted doe liveweights by mating season group and parity from 20 weeks prior to parturition up to 30 weeks postpartum, the approximate time at which rebreeding would be expected to take place under seasonal breeding management, are given in Figure 4.5. Note that estimates for synchronized week 20 and 24 correspond to predicted pre- and postpartum liveweights of dams. Least squares means and standard errors of liveweights by mating season and parity and percent changes therein at selected stages are presented in Tables 4.11 and 4.12. Liveweights of does at the time of conception did not differ substantially across mating seasons, except for group 6, in which does were markedly lighter than in all other groups. However, this result may have been biased due to the smaller number of observations in group 6 and due to the fact that a very largeproportion (72 percent) of the does joined in this group have had a live birth event within 40 weeks prior to mating. For comparison, figures for groups 1 to 5 amounted to, respectively, 26, 26, 47, 63, and 33 percent of the does joined. As expected, increases in liveweights during pregnancy, adjusted for the effect of parity, were largest (>32 percent increase relative to liveweight at conception) when mating took place at the beginning or just prior to the rainy seasons (groups 5 and 6), so that during most of the pregnancy stage animals were exposed to favourable forage conditions. Conversely, liveweight gains decreased gradually the further the pregnancy stage was displaced towards the long dry season (groups 1 to 3). At the measurement time immediately preceding parturition (synchronized week 20), the heaviest dams were observed in mating season group 1 (42 kg), and the lightest in mating season group 3 (38 kg), although differences in liveweight reached statistical significance only during the postpartum period. Weight losses between pre- and postpartum measurement times varied only slightly and ranged from 4.9 (group 4) to 6.4 kg (group 6). The average change in liveweight in the various groups during the suckling period (differences between week 24 and 38) ranged from a weight loss of 4.9 kg in group 1 to a weight gain of 2.2 kg in group 3.
|Figure 4.5. Estimated growth curves (least squares means) of does by mating season over a reproductive cycle of one year duration. The time origin corresponds approximately to the date at conception.|
Although differences across mating season groups in liveweights and the associated relative changes therein failed to reach significance during lactation, it is nevertheless clear that mobilization of body reserves was more pronounced when the lactation stage coincided with periods of low quality and quantity of forage on offer (groups 1 and 2) than with periods of good pasture availability (groups 4, 5, and 6). The response profiles in Figure 4.5 show that liveweight recovery in does joined in mating seasons 1, 2 and 3 was linked to the onset of the short rainy season, which occurred at a very late stage for group 1 (around week 46) and shortly after birth for group 3. Weight losses during early lactation were much smaller in groups 4 to 6, and at weaning time, slight relative weight losses could only be observed in group 4 (-0.9 percent). Does in mating season group 5 and 6 were still more than 12 percent heavier at weaning than at conception time. In response to the favourable feeding conditions prevailing throughout the long rains, a steady increase in body weights was observed in mating seasons 4 and 5 from 32 weeks onwards. Although does in mating season 6 were on average only about 0.5 kg lighter at weaning than at the postpartum measurement time, liveweight losses during the following long dry season were substantial, amounting to approximately 5.5 kg, such that animals were predicted to weigh slightly less after 52 weeks than at conception (-2.4 percent).
Except for mating season group 6, breeding does generally were able to fully replenish their body reserves by 52 weeks. The relative increases in liveweight of 3 to 12 percent (Table 4.13) in mating seasons 1 to 5 also indicate that, on average, immature females must have achieved net growth in body weight over the reproductive cycle. This is also apparent from the response profiles by parity class shown in Figure 4.6 and the corresponding estimated relative weight changes given in Table 4.13. First kidding does were significantly lighter at all measurement times, but also achieved much larger relative weight gains during the reproductive cycle than higher parity does. Relative weight changes decreased linearly (p <0.001) with increasing parity of does at all stages. While first kidding does on average had gained 2.9 kg (15.2 percent) by 52 weeks, third and higher parity does were on average 1.7 kg (-3.4 percent) lighter at that stage than at conception time. Does with one and two prior kiddings were on average 2.1 (8.4 percent) and 0.4 kg (1.1 percent) heavier after completing the reproductive cycle. From 28 weeks onwards, all pairwise comparisons of estimated relative weight changes by parity stage were highly significant.
Litter size exerted a marked influence on liveweight development of does, although this effect was not constant across parity levels (Figure 4.6 and Tables 4.13 and 4.14). At conception time, there was a slight tendency for higher liveweights in does carrying twin fetuses, the difference to does carrying a single fetus ranging from 0.6 kg (at least three prior kiddings) to 1.6 kg (two prior kiddings). Does carrying twin fetuses were not generally heavier at, and did not generally exhibit higher relative weight changes until the end of the gestation period than those carrying a single fetus. This was the case only in parity one and two does, which were predicted to be heavier by, respectively, 4 kg (10.7 percent) and 5.2 kg (13.2 percent) at 20 weeks from conception. The reverse was observed in nulliparous and third and higher parity animals, though here the estimated liveweight differences between those carrying single and twin fetuses were very small (1.3 and 1.8 percent, respectively).
|Figure 4.6. Estimated simple effects of litter size within parity class on relative liveweight change in does over a reproductive cycle of one year duration. The time origin corresponds approximately to the date at conception.|
Relative decreases in liveweights between pre- and postpartum measurement times varied only slightly across parity stages, and ranged from 10.9 to 11.8 percent in does that gave birth to a single kid, and from 13.5 to 16.4 percent in does with twin birth events. However, as may be seen from the response profiles in Figure 4.6, the depression in liveweights during lactation was much larger in does nursing twins, which in all parity stages exhibited net relative weight losses until weaning. This pattern was most pronounced in third and greater parity animals, with an average reduction in liveweight until weaning of 6.9 percent relative to that observed at conception time. In contrast, does nursing singles never exhibited net relative weight losses during the lactation stage except for parity class three and greater, in which dams were predicted to be about 6 percent lighter by the end of the reproductive cycle than at the time of conception. The sharp decline beyond 40 weeks in the latter curve of liveweight change, however, was related to both the small number of observations in this cell of the design from 44 weeks onwards (<15 observations) and to the absence of a control for environmental influences on responses in the fitted model. For instance, for ten out of the fifteen parity three or greater does nursing a single kid, the end of the lactation stage coincided with the long dry season.
Effects of sex, litter size, milk yield, and parity of dam
Consistent with expectation, sex and litter size exerted significant effects on birth weights and subsequent growth rates, with females weighing less than males, and twins less than singles. An interesting feature of the analysis is that relative growth rates, expressed as weight gain per day relative to average metabolic body weight, of female and male kids did not differ significantly from each other. This points to the fact that deviations in absolute growth rate (AGR) were entirely due to the larger birth and mature weight of males, and not to differences in growth intensity or maturing rate. Hence, upon expressing AGR in relation to average metabolic body weight status, size-dependent effects on observed growth performance tend to be reduced or eliminated. A similar conclusion could have been arrived at by using the relative growth rate measure (RGR) suggested by Fitzhugh and Taylor (1971), defined as the difference between the natural logarithm of two consecutive weight measurements divided by the length of time interval, which represents the average percentage change in body weight per day. As indicated by Fitzhugh (1976), RGR may also be approximated by AGR/[(y t+1+y t)/2], where y t+1 and y t are body weight measurements made at two consecutive time points; the only difference between this and the measure employed in the present study is that the denominator (average body weight) is not converted to the metabolic weight scale. Consequently, at any age, there was an almost perfect correlation (r>0.99) between the index used in the present analysis and RGR asdefined by Fitzhugh and Taylor (1971).
In contrast to the effect of sex, litter size affected both intercepts and shapes of growth curves, such that twins were not only lighter at birth, but also matured at a slower rate than singles until approximately 18 weeks of age, while the relationship was reversed thereafter. This pattern was undoubtedly due to the increased competition (at a given yield level) for milk in twin litters, given that the effect of litter size remained unaffected by milk yield level until weaning. Retarded growth caused by suboptimal levels of nutrition during the preweaning period implied that twin born kids were less mature at weaning than singles and this was probably conducive to the faster growth rate observed in surviving twins over subsequent time periods. However, this was not sufficient to fully compensate the difference in body weight, and singles were persistently heavier than twins by about 14 percent from one year of age onwards. Similar differences in body weights between twins and singles have been noted previously by Sacker and Trail (1966), who found single born male and female SEA Mubende goats to be heavier by 17 and 7 percent at one year of age, respectively, compared to their twin born contemporaries.
The observation that birth weights tended to increase with milk yield level until weaning appears to be largely due to the fact that milk yields themselves were found to be strongly correlated with liveweight of does at parturition time. Because of the presence of this collinearity, postpartum dam weight was not considered as a source of variation in analysing body weight development of kids. However, an examination of the effect of dam postpartum weight, adjusted for the effects of sex and litter size, revealed significant linear and quadratic trends (p <0.05) in birth weight with increasing postpartum weight of dam. This is in accordance with previous reports (Amoah et al., 1995; Devendra and Burns, 1983; Sherman, 1987), and with the general principle of productive output being proportional to the metabolic weight of the dam (Taylor and Murray, 1987). But postpartum weight itself may have been influenced by many factors, including degree of maturity, phenotypic variation in size, or nutritional state of dam during pregnancy. Hence, though postpartum dam weight certainly bears some relationship to maternal capacity, its use as a predictor variable probably does little to elucidate the mechanisms by which prenatal environment and maternal factors may affect birth weight of kids.
The effect of total milk yield until weaning on body weight gains followed the expected pattern, in that a clear linear trend in responses with increasing milk availability for youngstock was observed until about 10 weeks of age. This influence decreased steadily with advancing age and, by one year, differences in body weight were largely independent of the amount of milk available until weaning. It is interesting to note that, if they survived, kids which had access to the least amount of milk until weaning were able to compensate for this handicap by a phase of accelerated growth between 3 and 6 months of age. These increased rates of gain suggest that the smaller quantity of milk sucked was possibly compensated for by increased intake of pasture feeds at an early stage of life, so that such animals were effectively weaned and had fully developed rumens well before weaning at 4 months of age.
A curvilinear relationship was detected between dam parity and birth weight of kids, with first and second kidding does giving birth to the lightest and heaviest kids, respectively. At later ages, this pattern changed slightly in that markedly lower body weights were observed for kids born to fourth and higher parity does, relative to all others. However, parity did not affect relative or absolute growth rates, thus complicating the interpretation of the effect of this factor on body weight development of kids. The model used did not account for the effects of litter size and milk yield, since both were correlated with the effects of parity and mating season. Then, a possible explanation for the observed pattern is that the decrease in birth weights and growth rates associated with the higher probability of twin birth events in multiparous does was counteracted, to some extent, by an opposite effect exerted on these traits by the curvilinear trends in dam body weight and milk production with increasing number of lactations. Hence, although birth weights and growth rates could be expected to increase up to the third lactation due to concomitant increases in dam weight and milk yield, this may have been attenuated by the higher probability of producing twins in older does.
Effects of mating season
It is generally accepted that the rapid rate of fetal growth during the final stage of pregnancy imposes a metabolic strain on the doe (Dunn and Moss, 1992; Landau et al., 1996; Sauvant et al. 1991). When the dietary supply of nutrients is inadequate, these increased nutritional requirements are met by mobilising body reserves. Consequently, it has been stated that goats, like other domestic ruminant species (Dunn and Moss, 1992; Spitzer et al., 1995), are sensitive to nutritional stress during pregnancy, especially under extensive management conditions (Sibanda et al., 1997; Osuagwuh, 1992). Severe malnutrition during varying stages of pregnancy appears to exacerbate liveweight losses and increase the risk of reproductive wastage due to abortion, retardation of fetal growth, reduced birth weight, and increased neonatal death rate. Furthermore, periods of severe nutritional stress pre partum are likely to affect subsequent levels of milk production, thus further increasing the likelihood of neonatal losses and depressed growth performance.
In the present experiment, pronounced negative energy balances during gestation and, consequently, low birth weights, could be expected to arise in those treatment groups in which either the entire gestation period (mating season group 3), or the early (mating season group 4) or late (mating season group 2) stages thereof occurred during the long dry season, when feed quality and quantity were most limiting. Conversely, the more favourable feeding conditions prevailing between the months of October and April could be expected to lead to improved birth weights when gestation took place during this period, such as it was the case in mating season groups 5 and 6, as well as in group 4 from mid-pregnancy onwards. However, no systematic effect of mating season on birth weights of kids was detected. On the one hand, the large variation in birth weights accounted for by the random interaction effect of mating season with production cycle obviously contributed to the absence of a clear effect of mating season on birth weights. On the other hand, this was also in part due to the fact that litter size, which was correlated with mating season and parity, could not be included in the corresponding statistical model. Birth weights, however, were found to be significantly lower in twins than in singles. Hence, the low birth weight predicted for kids born at the onset of the dry season (mating season group 1) was certainly to some extent related to the larger proportion of twin birth events in this group (58 percent), which was well above that estimated for all other groups (< 48 percent). Inspection of the liveweight development profile during gestation indicated that does in mating season 1 gained on average 18.6 percent in weight until parturition, and exhibited the highest liveweights at mating and prior to parturition. On average, prepartum body weight losses did not occur, suggesting that fetal growth was probably not inhibited through nutritional constraints.
As expected, signs of severe nutritional stress during late pregnancy were apparent in mating season groups 2 and 3, in which does began to loose weight from about 6 weeks prepartum onwards, but this did not seem to have affected birth weights. Experiments previously conducted under controlled environmental conditions also failed to reveal significant effects of dietary treatments during late pregnancy on birth weights in goats. Sibanda et al. (1997) could not detect differential effects of low, medium, and high energy and protein diets fed 7 weeks before parturition to Matebele goats on birth weights. The authors point out that does in the low energy group could have compensated for nutrient deficiency by mobilizing body reserves to ensure development of the fetuses. In this group, net weight losses (about 5.6 percent) were only apparent when comparing liveweights at the onset of the dietary treatment in week 15 of pregnancy and the post-kidding measurement time, but did not occur between 15 and 21 weeks of gestation. In contrast, Osuagwuh (1992), working with West African Dwarf goats, found that underfeeding between 61th and 120th day of pregnancy significantly depressed birth weights and neonatal growth rates, while a reduction in the plane of nutrition during late pregnancy (>120 days) failed to produce similar results. Also, in a study conducted on Alpine goats, dietary energy concentrations ranging from 1.8 to 2.5 Mcal·kg— 1 DM fed from 90th gestation day onwards did not affect litter weights (Sahlu et al., 1995), suggesting that severe nutritional restrictions during late gestation may be required to deplete body energy stores enough to influence fetal growth.
From these results and from the review presented by Landau et al. (1996), it seems that the clear expression of differential effects of nutritional factors on fetal growth rate and subsequent birth weight is highly dependent upon the severity of under- or overnutrition, as well as upon the duration and the stage of pregnancy over which these changes in nutrient supply are experienced. Obviously, a further complication arises under field conditions in semi-arid climates, where nutrient availability at any time point during gestation is difficult to quantify, since it is not only determined by seasonal fluctuations in primary productivity, but also by random climatic perturbations which occur at varying time scales and may considerably alter seasonal patterns of forage production. This was partly reflected by the large variation in birth weights accounted for by the production cyclexmating season variance component (σ2 PC ⋅ MS=0.046), as opposed to the estimated variance of birth weights (σ2=0.015). Although non-significant, there nevertheless was slight evidence for a tendency of increasing birth weights when gestation took place between October and April (mating seasons 5 and 6), which generally corresponds to the most favourable period in terms of both feed quality and quantity. Inspection of the liveweight development profiles of does from mating to parturition, adjusted for the effect of parity only, revealed that the largest relative liveweight gains during gestation were achieved in the latter two treatment groups (32.2 and 37.8 percent, respectively), as opposed to all other groups (<21 percent).
The experiment succeeded in demonstrating that significantly different growth rates achieved by kids could be attributed to mating season, and hence season of birth. The range of absolute growth rates among mating season groups of 44 to 57 g·day— 1 from birth to 48 weeks of age estimated in this study were somewhat higher than the 38.3 g per day from birth to one year of age observed for SEA goats under comparable husbandry conditions by Wilson et al. (1984). The marked differences in growth patterns across mating season groups may largely be explained on the basis of availability of pasture, particularly during the preweaning period. This conclusion is supported by the finding that both range condition scores had significant impacts on estimated relative growth profiles. The highest average preweaning weight gains were achieved by kids born in the period from October to May (mating season groups 3 to 6), while growth [page 88↓]performance was seriously compromised when birth took place at the middle (mating season group 1) and, to a lesser extent, toward the end of the long dry season (mating season group 2). Growth performance until 3 months of age was severely depressed in the former group as pasture conditions deteriorated in the dry season and does could not cover their nutrient requirements for sustained milk production.
When compared on the basis of relative growth rates, the growth performance estimated from the present data is in agreement with that obtained previously for SEA goats by Wilson (1958) until 8 weeks of age (Table 4.15), though in his study kids appeared to be markedly heavier at birth. After allowance has been made for differences in mature size, it can be seen from the same table that comparable early growth rates are also achieved by other small-sized tropical goat breeds. Comparing growth performance among and within breeds and strains on the basis of absolute growth rates clearly is not very meaningful if there are substantial differences in mature size. Generally, absolute growth rates are highly correlated with body size at maturity, which bears no direct relationship with higher biological efficiency (Brown et al., 1976; Dickerson, 1978; Ogink, 1993). For instance, a large European goat breed such as the German Fawn achieves much higher absolute average daily gains than all other breeds listed in Table 4.15. However, most of this difference is probably due to the variation in genetic size at maturity, as indicated by the performance of the German Fawn in terms of average daily gain relative to metabolic liveweight state. When compared on the basis of this criterion, the latter breed does not appear to be superior to the smaller-sized tropical breeds. In fact, German Fawn goats even seem to grow at a slightly slower relative rate than the others over the first weeks of life. Although it cannot be ascertained based on the figures presented in Table 4.15, such a trend would most likely be an expression of the well known negative genetic relationship between size at maturity and earliness of maturing (Fitzhugh and Taylor, 1971; Taylor and Fitzhugh, 1971; Brown et al., 1976; Fitzhugh, 1976). This negative relationship implies that individuals growing to the heaviest mature weight tend to be less mature at a given age or older at a given degree of maturity.
After weaning, environmental conditions continued to have considerable effects on body weights and weight gains over short time periods of time. However, compensatory effects tended to reduce the net effect of environmentally induced fluctuations in availability and quality of nutrition on estimated weight-at-age profiles. Thus, while substantial differences in body weights could be observed at weaning time among mating season groups, these had almost completely disappeared by one year of age, with the exception of kids belonging to mating season group 6. In the latter group, kids weighed on average 9.6 (±0.35) kg at weaning, at which time they passed through a period of severe nutritional stress coinciding with the middle of the long dry season. During the subsequent period from October to May, which may be considered a re-alimentation phase, compensatory gains were noticeably lower than, for instance, in mating season group 1, in which kids were younger and at a lower degree of maturity. Specifically, at the start of the short rainy season in late October, kids in mating season group 1 were on average 16 weeks old and were estimated to weigh 7.9 (±0.35) kg, while those in mating season 6 were on average 24 weeks old and had an average weight of 10.4 (±0.48) kg. The magnitude of compensatory gains have been reported to be highly dependent upon the physiological maturity at the time of re-alimentation after a period of nutritional stress. In a review on this subject, Hogg (1991) concludes that the potential for the greatest compensatory response apparently occurs in animals whose weight is near 25 to 30 percent of mature size, while above and below this weight their potential to respond declines. Kids in group 6 appeared to be close to the upper end of this range at the time of re-alimentation, and this may partially explain their lower response.
Based on the growth and weight-at-age profiles for mating season 6 youngstock (Figures 4.1 and 4.4), it seems that nutritional restriction may have induced a reduction in mature size. However, Ryan (1990) argues that the level and duration of nutritional restriction required to depress mature size may be far more pronounced than those generally associated with the seasonal variations in pasture quality and quantity on offer. Alternatively, it is possible that animals in group 6 would have continued to grow at the slow rates observed during the second year of life and, eventually, would have reached a similar mature weight as those in other groups, though at a much older age. However, these conclusions must remain tentative, since the small number of observations beyond 64 weeks of age clearly affected the estimation of growth rates and weight-at-age curves. For instance, while weight losses were predicted to occur between weeks 88 and 104 in group 6, this was not reflected in the corresponding weight-at-age curve, which indicated increasing weights from 88 to 104 weeks of age.
Effects of litter size and parity of dam
As expected, nulliparous dams were still growing and achieved much higher relative weight gains throughout the reproductive cycle than higher parity dams. The estimated liveweight development profiles by parity class indicate that weight gains levelled off approximately at the end of the third lactation, at which stage dams weighed about 36 kg. The average weight range of 27.7 (mating season 6) to 36.9 kg (mating season 1) observed at the end of the reproductive cycle in the present strain of SEA goats appears to be markedly higher than that reported previously by Ruvuna et al. (1991) for the same breed in Western Kenya. In their study, liveweights at 5 months postpartum ranged from 26.4 kg in two years old does to 30.76 kg in 6 years and older does. Similarly, Wilson et al. (1983) reported postpartum weights in Maasai SEA goat flocks ranging from only 23.4 kg in first breeders to 28.9 kg in fourth parity dams.
Relative liveweight increases during gestation were not generally larger in twin than in single carrying dams. Such a trend could only be detected in second and third parity does. The assumption is made that nulliparous and old animals (at least three prior kiddings) were less able to withstand the increased physiological stress associated with carrying twin fetuses to term, and therefore exhibited a slower rate of liveweight gain than their contemporaries carrying a single kid. In addition, the observed pattern may also be the expression of a reduced rate of body weight recovery following periods of nutritional stress, or of a lower rate of accumulation of body reserves during pregnancy. Within parity classes, the differences in liveweight due to litter size are most evident when comparing weight estimates obtained for week 20 of the reproductive cycle, i.e., just prior to parturition. For nulliparous and old does, no difference in body weights due to litter size was apparent, while for parities 2 and 3 twin carrying does were significantly heavier than those carrying a single kid. Similarly, mobilization of body reserves during the early stages of lactation appeared to be much more pronounced in the former than in the latter.
Effects of mating season
The different patterns of nutritional conditions engendered by the mating season treatment produced marked differences in liveweight changes and weight development curves in does over a reproductive cycle of one year duration. With respect to liveweight development during gestation, does were clearly advantaged when they were mated just prior to the long rainy season (mating season 1). Both liveweight at mating and at the prepartum measurement period were found to be heaviest in this group (35.4 and 42.0 kg, respectively). On the other hand, the largest relative liveweight gains of 32.2 and 37.8 percent in relation to liveweight at conception were achieved by does mated from October to January (mating season 5 and 6, respectively). A large part of these weight gains went into the replenishment of body reserves that were depleted during the preceeding long dry season.
When considering relative liveweight changes, net liveweight increases over the gestation period can be assumed to have occurred at least in those groups for which markedly positive relative liveweight changes were estimated until the postpartum measurement time point. This was the case in mating season groups 1, 4, 5, and 6 in which dams were, on average, heavier by 10, 13, 22, and 25 percent, respectively, at synchronized week 22 than at conception time. In a study on Scottish Half-Bred and Dorset Horn ewes, Treacher (1970) reported that in order to prevent utilization of fat reserves, relative liveweight gains of about 20 and 22 percent were required during late pregnancy. Over the last eight weeks of gestation, gains of this magnitude could not be observed in the present study. Thus, the estimates obtained from this data suggest that, to a large extent, increases in dam liveweight during gestation could be attributed to the growth of conceptus and/or the replenishment of body reserves. In contrast, nutritional stress during late pregnancy apparently had an adverse effect on dam liveweights in mating season groups 2 and 3. The respective liveweight development curves showed that goats had begun to lose weight from about 6 weeks prepartum onwards; this was undoubtedly related to the deficiency in the quality and quantity of feed resources at the peak and toward the end of the long dry season.
Consistent with expectations, the increased nutrient requirements at the onset of lactation generally could not be covered by available feed resources, so that this stage of the reproductive cycle was accompanied by massive mobilization of body reserves. Losses in body weight were especially drastic in females which gave birth at the start and at the middle of the long dry season (groups 1 and 2). In the latter groups, the peaks in average body weight losses relative to postpartum weight (18.4 and 11.8 percent in groups 1 and 2, respectively) were comparable to that observed by Mbayahaga et al. (1998) in a local Burundian strain of SEA goats after parturition in the dry season (11.1 percent). The fastest rate of recovery was observed in mating season 3, in which dams began to gain weight as soon as 5 weeks postpartum, in response to improving nutritional conditions at the onset of the short rainy season. At the other extreme, females which [page 90↓]gave birth at the beginning of the long dry season (mating season 1) were severely affected by the nutritional restriction experienced during the first four months of lactation, and these animals continued to lose weight until pasture forage production had improved in response to the onset of the short rainy season. Overall, liveweight development patterns during lactation were very different from those described previously for an extensive goat husbandry system operating under Mediterranean climate (Zygoyiannis and Katsaounis, 1986). The evidence presented in the latter study showed a steady increase in both liveweight and body condition score immediately after parturition and throughout lactation. The authors concluded that the available nutrition was at least adequate for the requirements of the indigenous breed studied. In the present setting, however, apparently in none of the mating season groups was the nutritional environment such that goats could display similar patterns of liveweight recovery after parturition. On the other hand, Gall (1980) supports the view that substantial mobilization of body fat stores at the beginning of lactation is a phenomenon particularly marked in goats due to their high initial milk production, for which the required dietary energy cannot be ingested by the animals even when fed on concentrates.
Under similar climatic conditions in Mali, Wilson and Light (1986) found that birth in the post-rain period and during the long dry season immediately following it had a negative impact on the reproductive potential of West African Sahel goats. The authors argued that due to poor nutrition such females took a longer time to return to breeding status, while does that kidded in the short rainy season recovered at a much faster rate due to the better forage availability. The present data, however, did not reveal comparable effects of liveweight development of does on their reproductive performance. In spite of the fact that animals allocated to mating season 1 lost, on average, about 17 percent of their body weight over the period from parturition to 22 weeks of lactation, no adverse effects on subsequent fecundity rates could be detected. During the last 10 weeks of the reproductive cycle relative liveweight gains of about 18 percent were achieved and does were estimated to recover at a rate roughly proportional to their weight at conception time. Similarly, the rapid decline in liveweight toward the end of the reproductive cycle observed in mating season 6 (-17 percent) apparently did not affect reproductive performance of does.
Seasonal breeding does not appear to confer any major advantage in terms of liveweight production of youngstock, except for the fact that mating during the short dry season from December to January should perhaps be avoided in order not to compromise liveweight development of kids. Wilson et al. (1984) arrived at similar conclusions in studying the effects of season of birth on liveweight development of kids in Maasai SEA goat herds. They found no significant differences weights of 18 months old kids and concluded that there would be little to be gained in terms of growth performance from attempting to control breeding. Nonetheless, birth weight and liveweight development of kids during the preweaning period are factors which can be expected to influence overall herd productivity through their effect on survivability. Indeed, as revealed by the present data, both survivability until weaning and weaning weights were highest when kids were born over the period from October to April (mating seasons 3 to 5).
Although body weight development profiles of does were seemingly dissimilar contrasted across mating season groups, this did not result in significant differences in estimated body weights of does at the end of the reproductive cycle, as long as goats were joined over the period from February to November (mating seasons 1 to 5). It is also worthwhile to note that within the latter groups there was virtually no difference in estimated weight at conception time and at 52 weeks after adjustment was made for the effect of parity. Mating in December and January (mating season 6), however, clearly resulted in markedly lower liveweights at the beginning and end of the reproductive cycle. In terms of balancing nutrient requirements of the dam and nutrient supply from pastures, this breeding period can be considered the least favourable among those considered in the present experiment.
|© Die inhaltliche Zusammenstellung und Aufmachung dieser Publikation sowie die elektronische Verarbeitung sind urheberrechtlich geschützt. Jede Verwertung, die nicht ausdrücklich vom Urheberrechtsgesetz zugelassen ist, bedarf der vorherigen Zustimmung. Das gilt insbesondere für die Vervielfältigung, die Bearbeitung und Einspeicherung und Verarbeitung in elektronische Systeme.|
|DiML DTD Version 3.0||Zertifizierter Dokumentenserver|
der Humboldt-Universität zu Berlin