The purpose of the baseline run is to establish the validity of the model for the system under study. Using the production parameters as described in the previous section and applying the management practices presently followed in this system, 10 replications over a 15-year period are made and the simulated performance of a herd over this period is compared with data obtained from the real system. Each replication is assumed to start with a "good" year type (see Table 1), and thereafter year types are drawn probabilistically. Appendix Table A.6 presents the simulated year type sequences for the 10 replications and the overall occurrence of each year type. Simulated year type frequencies correspond closely to the assumed probabilities of occurrence in Table 1.
Validation of the model for this production system involves comparison of field data with simulated results. Real input and output data are available for the system under study over a 10-year period from 1970 to 1980 (APRU, 1980a and 1981)14. The model is considered to be validated for the production system under study if the following criteria are satisfied:
a. the simulated values of key production traits correspond closely with values from the real system; andb. the simulated growth curves for indicator classes of livestock correspond closely with recorded performance.
For model validation under criterion (a) five key production traits are considered:
1. The calving rate, i.e. the number of calves born divided by the number of breeding females at the end of the breeding season (early April), adjusted for interim sales.2. The survival rate of calves up to 12 months of age, i.e. the number of 1-year-old calves divided by the number of calves born.
3. The survival rate of animals up to 2 years of age, i.e. the number of 2-year-old animals divided by the number of calves born.
4. The weaning weight, i.e. the average weight of all calves at 7 months of age.
5. The 18-month weight, i.e. the average weight of all animals at 18 months of age.
14 On the input side the 10 actual years involved, classified on the basis of annual rainfall (although annual rainfall alone does not adequately reflect forage quality), correspond to 3 "good", 2 "above average", 2 "average", 1 "below average" and 2 "poor" years. Their corresponding frequencies are 0.1, 0 3, 0.2, 0.1 and 0.2, compared with the simulated frequencies of 0.167, 0.287, 0.33, 0.173 and 0.04, respectively (Appendix Table A.6). It would have been a coincidence if the frequency distribution of the short 10-year period had closely matched that of the longer term. However, a comparison between the average and expected outcomes is more favourable. The average outcome of the actual 10-year period falls halfway between an "above average" and an "average" year, which is very close to the expected outcome, based on the longer term probability distribution, which falls two thirds of the way between the same year types.
The comparison between simulated and actual values of the above five production traits is presented in Table 4. The simulated values correspond very closely to the actual ones, with a maximum divergence of 1.5%.
For model validation under criterion (b), actual detailed liveweights exist for T animals only, and only up to 2.5 years of age, so a comparison of actual and simulated figures is possible for that genotype only. Figure 5 shows simulated average seasonal liveweight fluctuations of T females from birth to maturity, together with corresponding actual values (to the extent available) from the real system. Within the extent of available data from the real system, simulated liveweights correspond closely to actual liveweights, in terms of both seasonal liveweight changes and the overall trends with increasing age.
After weaning (May to July) there is a drop in simulated liveweights of about 30 kg during the dry season and animals reattain their weaning weight about 6 months later during November to January. Thereafter the fluctuation in their liveweights follows the seasonal variability of forage on offer. After maturity, at about 480 kg which is reached between 4 and 5 years of age, cows lose on average 100 kg during the dry season, which they regain during the wet season. In absolute terms, simulated liveweights for mature cows fluctuate on average between 425 and 525 kg 15.
15 Note that the simulated liveweight fluctuations reported in Figure 5 derive from simulations including the full range of year types. As such they indicate seasonal liveweight ranges which should be expected to be wider than would occur in a normal forage year, i.e. an average year type. Similarly, the seasonal liveweight fluctuations which would occur in the real world in a very poor forage year would be likely to cause wider ranges in liveweight to occur than those simulated in the baseline run.
Table 4. Comparison between baseline simulation results and actual production for T and ST herds.
|
Production trait |
T |
ST |
||
|
Actual a |
Simulated |
Actual b |
Simulated |
|
|
Calving rate (%) |
87.5 |
88.5 |
91.9 |
91.8 |
|
Survival rate to 1 year (%) |
93.5 |
94.3 |
94.8 |
94.5 |
|
Survival rate to 2 years (%) |
92.5 |
92.8 |
93.8 |
93.0 |
|
Weaning weight (kg) |
194 |
191 |
216 |
218 |
|
18-month weight (kg) |
320 |
322 |
355 |
351 |
a Actual values for T are from APRU (1980b; 1981).
b Actual values for ST are inferred from comparisons between indigenous and crossbred cows at APRU's experimental station at Musi (APRU, 1980b; 1981). These results indicate that ST animals perform considerably better than T; calving rates were about 5% higher, survival rates about 1.4% higher, and weaning and 18-month weights about 11 % higher.
On the basis of these criteria and the subjective evaluation of the simulated results by field personnel familiar with the production system, the model is considered to replicate the real world with an adequate degree of accuracy for practical purposes.
Figure 5. Comparison between simulated and actual liveweight evolution curves for Tswana females.
