The long-term survival of traditional livestock production systems within the rapidly evolving national economies of Africa will depend on their capacity to provide products in quantities and at prices which satisfy the subsistence and income needs of the livestock producers. Equally, the survival of these production systems will depend on their impact on land resources, as existing surplus capacity can quickly be eliminated by poor management or over-use. Human populations in the main livestock producing areas of Africa are increasing, and there is a strong positive linkage between the sizes of human and livestock populations due to the need for stock to meet subsistence needs. Thus these systems will be expected to produce even more in future from the same land resource base.
Existing pastoral systems evolved over many years when the pressure on grazing lands was less intense than it is at present. Thus the limitations and fragility of the resource base were not appreciated, either by the pastoralists themselves or others concerned with their welfare. Periodic catastrophes caused by drought or disease were expected, with their attendant impacts on animal and human populations. This situation has changed somewhat over the last few decades, with the introduction of a variety of technical innovations and structural changes. For example, the rinderpest campaign has removed a major source of mortality in domestic cattle across the continent; however, by permitting livestock producers to keep higher average herd sizes over time, the elimination of rinderpest has indirectly contributed towards increasing pressure on grazing resources. The development of water supplies has made grazing areas accessible which were previously little used and changed seasonal grazing patterns. Overall, this development has been favourable, but it has also resulted in overgrazing and non-reversible range degradation in many areas.
Certainly many of the effects of development, both positive and negative, were foreseen, and, on balance, the changes brought about have probably been positive. However there are now few relatively empty areas in the pastoral zones of Africa and pastoral production systems are, according to many observers, already near the limits of their capacity. For this reason, it has become increasingly important that the consequences of any proposed changes be fully elaborated before they are introduced to the real systems.
Despite important differences among various livestock production systems in Africa, for example in the degree of association with agriculture or the extent and nature of stock movements, all are concerned essentially with the conversion of forage and other feed resources, via the animals and in the context of management regimes, into flows of products of use to man. The management regimes regulate the scheduling of some of the biological processes (such as the breeding season and age of calves at weaning) and the timing and rates of offtake from the herds.
The forage resources of many African pastoral systems are held communally. They are characterized by low levels of productivity per unit area and high variability in yields, both within and across years. Where grazing resources are held in common, the individual herd manager has few choices or opportunities to improve the supply of forage to his herd at any given time, and is limited primarily to moving his animals as the forage resources in one location are depleted. Some systems similar to commercial ranching do exist, with privately owned stock on privately owned and fenced land, as for example in Botswana and Kenya. However, they account for only a small fraction of the total yield of livestock products from the pastoral areas. In both types of production system, there is a substantially high residual variability in the resource base, and this is becoming more severe as livestock and agricultural enterprises expand, restricting the range of herd movements. This variability in the supply and quality of forage on offer compounds the variability inherent in the animal-level biological processes.
However, within the confines of the production environment, managers have important means at their disposal to regulate their herds and the flow of products from them. To a large extent the day-to-day or season-to-season performance of the herd is determined by the management practices followed. The benefits of some of these practices are obvious, but other practices have complex multiplicative and feedback effects which may become apparent only after considerable time. For example, in a particular production system taking a higher proportion of cow's milk for human consumption may be considered desirable. The uninformed observer, judging from a seemingly small short-term effect on calf growth, may be inclined to recommend such a practice on a routine basis. Such a decision, however, would ignore what might be a major effect on the condition of the cows and their subsequent reproductive performance, as well as the long-term effects on the calves of energy deprivation at a crucial stage in their development. These effects are not apparent at the time when the extra milk is taken. Another example is late weaning of calves. If the females of a breed have relatively long and highs-yielding lactations, any extension of the calf weaning age will make extra energy demands on the dams. If this extended lactation period coincides with the season when forage is limited and of poor quality, the cows will lose considerable weight and the period to the next conception will be delayed, resulting in lower calving rates. The gains to the calf from the additional milk may or may not compensate for the cost in reproductive performance. Again, the tradeoffs are not clear.
These simple examples illustrate the complexity of the interactions between a cattle herd, the forage resource and the management regime, and the importance of having a facility to make forward projections of the consequences of changes in a cattle production system. Such a facility, an analytical model, should be able to project herd performance and productivity, taking into account the essential responses of animals to forage and management practices.
A model of cattle production systems will have several features if it is to be applicable to a range of production situations and of use as a tool to complement more traditional research, particularly by allowing the integration of research results into a more holistic view of the target system. These features are described below:
1. The processes of reproduction, growth and death are fundamental to cattle production systems: together they are the major determinants of overall herd productivity. As such, they should be depicted separately in the model and their effects should, to the extent possible, be determined endogenously as functions of other factors embedded in the model.2. As there are many components of livestock systems which are inadequately researched and understood, the more complex and comprehensive a model the more appropriate it is to incorporate stochastic features reflecting the limited understanding of the processes involved. This implies that some of the components of a herd model will be stochastic because the determinants of many processes cannot be specified accurately in detail. For example, the complex set of variables which determine forage supplies to a herd are incompletely understood, as well as sufficiently complex to - require a major separate modelling effort. Thus forage supplies in a herd-level model are best represented probabilistically (i.e. as stochastic variables), thereby reflecting the observed variability in real production systems.
3. A model must be validated if it is to be useful for studies of a particular production system. Thus, the structures in a model should be based upon data readily observable in the corresponding real systems. Similarly, the output of a model should principally be parameters with real-world counterparts. The use of artificial variables should be minimized because the functioning of any component of the model with artificial variables can only be verified and validated as a logical expectation to which no probability can be attached.
4. The model must be time dynamic in order to schedule correctly events and the responses to events parallelling those in real production systems. The computational cost of specifying time in a model as a continuous variable is massive. Thus a discrete time-step must be specified for calculation, having regard to computational requirements, the availability of data for model specification and validation, and the appropriate interval to account for the important dynamics in the real systems.
5. Models which represent animals as integer entities have a direct correspondence to reality, as in such models animals are born, die, are pregnant or not pregnant, and so on. The mechanisms in the model then manipulate the status of individually simulated animals. This has operational advantages over non-integer formulations as computation are generally more efficient and the output is readily understood. In non-integer formulations, a herd is described by classes of animals which are to a large extent artificial. These classes vary through time, even for the same herd size, as they are fractionated and recombined. The interpretation of output from models with non-integer representations of animals is more difficult, particularly for small herds, which is usually the case in Africa.
6. Where a variety of different management regimes are feasible, and this is the rule rather than the exception in African cattle production systems, the model should permit specification of regimes in sufficient detail to allow simulation of the available options.
7. Optimizing models oblige the analyst to specify an objective function. As herds in Africa are managed to satisfy many objectives, and the ranking of these often varies through time according to the status of the system, it is more appropriate to formulate and use non-optimizing models of livestock production systems. The output from such non-optimizing models can then be evaluated relative to a range of objectives. This approach avoids a common problem of drawing incorrect conclusions because of mis-specification in the objective function.
8. To be useful for the analysis of a production situation other than the one for which it was first developed, a model should be designed so that its components can be modified, added to or deleted with a minimum effort. A modular structure is best able to satisfy this requirement. It is also important to document a model fully so that it can be used by persons other than those originally responsible for its design and development. Finally, the computer code should be written in a language routinely available on a wide range of digital computers - for example FORTRAN.
Together, these features imply that the best model of a livestock production system is time dynamic, stochastic, non-optimizing and treats simulated animals as individual entities. Several different models of cattle production systems have been developed for particular purposes over the last decade, but none has included this combination of features. Most have emphasized economic analyses, with insufficient attention to the underlying biological processes. By contrast, most of the changes introduced into livestock production systems in Africa have been technical innovation packages focused primarily on increasing biological performance; only rarely have development efforts focused exclusively on altering the economic context of production.
The World Bank developed a herd projection model (IBRD, 1972) which has been followed by a number of similar models (such as BAE, 1974; IADB, 1975) which simulate future herd numbers in yearly time-steps, with deterministic herd productivity parameters for each year provided as data. The IADB (1975) model is a derivative of the IBRD one (1972). These models are most often used to calculate the rates of return for various sets of assumptions about the effects of investment. The BAE (1974) model is similar in most respects, but also includes a faculty to preset the sequence of year-types occurring throughout the planning period. The different year-types assign different carrying capacities to the range area being considered. In this way, this model represents an important source of variability in the system, albeit simplistically.
The ranch model developed at the Centro International de Agricultura Tropical (CIAT) (Juri et al, 1977) is of more general application. It also uses a yearly time-step and stresses the financial aspects of ranch development, but, in addition, it has several stochastic components which permit evaluation of the risks involved in pasture development. It was designed for application in a ranching area of eastern Colombia where pasture establishment has a significant chance of failure in any year.
Dahl and Hjort (1976, 1979) developed a simple projection model to evaluate long-term herd dynamics on the basis of different assumed calving and mortality rates. This model focuses on aggregates and uses only two parameters to describe herd performance, and therefore can be used only for understanding demographic trends in large populations of stock. The authors have used the model to evaluate the recovery period for cattle, camel, and sheep and goat herds from a hypothetical 2- or 3-year drought.
Researchers in the Animal Science Department of Texas A&M University (Sanders and Cartwright, 1979; Smith, 1979) have developed, tested and applied a cattle production model which focuses on biological responses at the herd level to various sets of production conditions. The specification of a cattle genotype is provided as data to the model. It is a deterministic model and comparisons between runs are made using the results generated when steady-state conditions are reached at a future date in simulated time. This model is an important advance over others which do little to represent the basic biological processes. Furthermore, calculations are made in monthly time-steps. In the original formulation of the model, however, cattle in the simulated herd are not represented as integer entities, which makes it more appropriate to systems where the herding units are relatively large. Development of this model is continuing, including an integer version, and the perceived utility of its application to production problems in Guyana (Davis et al, 1976), Colombia (Cartwright et al, 1977), Venezuela (Ordonez, 1978) and Botswana (ILCA, 1978) has encouraged the group to initiate work on a sheep and goat model with similar features.
The structure of the model described here was influenced by ILCA's experience in the application of the original Texas A&M model in Botswana (ILCA, 1978) and the features considered essential for the evaluation of livestock production alternatives in Africa as discussed above. Chapter 2 provides the analytical background to the biological relationships represented in the model, based on quantitative evidence reported in the literature. The detailed algorithms used in the computer simulation model are elaborated in Chapter 3. Chapter 4 outlines the necessary steps for the application of the model. The appendices suggest analytical procedures for estimating parameters required as inputs to the model, based on field data.
