Previous Page Table of Contents Next Page


Effect of disease control programs


Modelling vector-borne disease epidemiology and the impact of control programs
Needs for modelling socioeconomic and environmental impacts of livestock disease control
The relationship between infections, diseases and their economic effects
Modelling livestock productivity
Potential for modelling ecological responses to the control and prevention of disease in African livestock populations
Session discussion


Modelling vector-borne disease epidemiology and the impact of control programs

B.D. Perry

International Laboratory for Research on Animal Diseases
P.O. Box 30709
Nairobi, Kenya

It is anticipated that improved control of tick-borne diseases and trypanosomiasis will be made possible through the development of recombinant antigen vaccines, enhanced genetic resistance and other technologies, the application of which will need to be responsive to the varying demands for disease control in different regions and production systems of Africa. The demand for control of tick-borne diseases and trypanosomiasis varies considerably in the continent depending on their distribution, the level of losses caused by them and on the economic outputs of the livestock production systems in which they occur. Given these variations, and the impracticality of gathering data representative of all possible conditions, the use of models offers a strategic method of quantifying the productivity effects of disease, permitting socioeconomic impact evaluations of disease control measures to be performed.

The level of losses caused by tick-borne diseases is affected by numerous factors, particularly host susceptibility, the dose of infection (dependent partly on tick infestation levels) and the age at which infection occurs. Thus in cattle production systems with indigenous East Africa zebu in the Lake Victoria Basin, where at least two generations of the vector Rhipicephalus appendiculatus occur each year, where all instars may occur on livestock at the same time, where the majority of animals are carriers of Theileria parva at low levels of parasitaemia, and where virtually all animals are infected as calves and become immune before reaching three months of age, little or no clinical East Coast fever (ECF) occurs as a result of T. parva infection. In contrast, in areas where R. appendiculatus infestations are strictly seasonal and only one generation occurs each year, where clinical cases (with higher piroplasm parasitaemias) occur, and where all animals are not infected as calves, even indigenous cattle may experience outbreaks of clinical theileriosis. In both areas, the introduction of Taurine cattle and their crosses increases the incidence of clinical disease.

The situation in which little or no losses occur due to clinical disease, termed endemic stability, only exists for T. parva infections in a few areas of eastern Africa, but for babesiosis and anaplasmosis it is much more widespread in the continent. The situation for heartwater is not well documented, but widespread endemic stability is suspected in many areas. The artificial induction of endemic stability through the use of vaccines will provide the most effective and sustainable option for tick-borne disease control in the future. The determination and quantification of the variables contributing to endemic stability and instability are therefore crucial in identifying target populations for tick-borne disease control programs, and assessing their impact.

The most important quantitative indicators of the presence of endemic stability and instability to tick-borne diseases are incidence of infection, incidence of disease, case-morbidity and case-fatality rates in young cattle. Stability is characterized by a high incidence of infection in this age group, but low levels of disease. Instability is generally characterized by a low and variable incidence of infection, and high incidence of disease. Regrettably, disease incidence, case-morbidity and case-fatality rates are rarely measured accurately under field conditions, and estimates of disease occurrence generally rely on antibody prevalence rates as a surrogate for incidence of infection (and for prevalence of immunity). Thus, it is important to determine whether it is possible to quantify the relationship between antibody prevalence and these indicators under different conditions and with different tick-borne disease combinations, and to determine the relationship between these indicators and productivity effects (in terms of milk, meat, traction and manure). An important component of this process is to determine the relationship between antibody prevalence, as measured by current and developing serological tests, and population immunity in the sampled cohort to the spectrum of tick-borne disease antigens they are likely to encounter.

A first step in modelling this dynamic process has been made for T. parva by Medley, Perry and Young (described earlier in the workshop), who simulated endemic stability and tested the effect of tick abundance and carrier state prevalence on its maintenance. However, in order to assess the validity of such models on a broader scale, they require testing in other endemically stable and unstable states. Regrettably, due to the intensive nature and high cost of prospective studies needed to validate such models, few data sets exist.

Eventually, it is hoped that such studies will lead to the development of user-friendly models that assess the efficacy of tick-borne disease control options for given herds, districts or regions, determine their effects on livestock productivity and assess their economic impact. With tsetse-transmitted trypanosomiasis, similar models are required, but due to differences in the immune mechanisms involved, it may not be possible to base them on the concepts of stability and instability. Furthermore, it is anticipated that the mechanisms involved are more complex and less predictable.

Needs for modelling socioeconomic and environmental impacts of livestock disease control

A.W. Mukhebi

International Laboratory for Research on Animal Diseases
P.O. Box 30709
Nairobi, Kenya

One of the research areas of ILRAD requiring modelling is the assessment and prediction of the probable economic, socio-cultural and environmental impacts of alternative control technologies for trypanosomiasis and tick-borne diseases in different production systems and agro-ecological zones in Africa and elsewhere. The application of new technologies to control these and other diseases may cause unforeseen economic, social and environmental consequences for segments of the human population. These effects, which may be better understood and projected through modelling, can include changes in livestock and human population densities, changes in patterns of land use accompanied by environmental changes, increased conflict over resources, disruptions of local social patterns, and increases in income and wealth disparities within the population. Moreover, other constraints within the overall livestock management system may serve to suppress potential benefits resulting from improved livestock disease control. These can include the existence of other diseases, poor animal nutrition, labour scarcity and lack of access to supporting services such as animal health delivery system, artificial insemination or credit for improvement of livestock. Modelling can help in identifying such constraints, or in evaluating their effects on potential impacts of alternative control strategies. Ex-ante assessment and prediction of the impacts of disease control through modelling can also help in avoiding possible deleterious effects in the application of new control technologies. It also provides useful information to farmers, governments and donors in order to assist them in planning better (strategic and systematic) disease control programs. In addition, such information can help in prioritizing the allocation of resources to disease research.

There are not always sufficient data, in content or quality, for assessing and predicting socioeconomic and environmental impacts of disease control. In many instances, the gathering of such data by field studies is unrealistic and costly, particularly given the large and varying geographic areas involved. A more cost-effective approach is to develop models using available field and secondary data that are supplemented by expert opinion, and then use such models for desk-top experimentation and extrapolation in ex-ante analyses. An additional advantage of models is that once developed and verified they provide a standardized method of data analysis which can be extended to users in different areas and countries, particularly if the models are developed and packaged in user-friendly computer software.

ILRAD has developed and applied spreadsheet models for assessing annual economic losses due to East Coast fever and the financial and economic impacts of its control by immunization at farm and above farm level. These models are being adapted and improved, in collaboration with the AP Consultants of the UK, for the analysis of trypanosomiasis control. In collaboration with Texas A & M University, a whole farm simulation model has been developed and applied for assessing and projecting farm level financial and nutritional impacts of ECF immunization under small-holder dairy production systems; a similar simulation analysis approach can be extended to trypanosomiasis control in sedentary livestock production systems. ILRAD's current modelling efforts, however, tend to emphasize microfinancial/economic aspects of livestock disease control effects. Macro-economic sectoral, trade and policy issues are not fully incorporated. In addition, linkages of financial/economic relationships to physical (ecological), biological (epidemiological) and socio-cultural (welfare) factors of livestock production systems are weak. Future modelling needs will require integration of micro- and macro-economic factors involved in livestock disease control with ecological, epidemiological and socio-cultural factors for more comprehensive assessments of the impacts of alternative control strategies. Such integrated model(s) will, however, be complex, and technical, data procurement problems and modelling mechanisms required are issues that must be addressed. The model(s) developed would ideally be suitable for application to different livestock production systems in different environments.

The relationship between infections, diseases and their economic effects

R.S. Morris* and W.E. Marsh+

* Department of Veterinary Clinical Sciences
Massey University
Palmerston North, New Zealand

+ Department of Clinical and Population Sciences
University of Minnesota
St Paul, Minnesota, USA


Abstract
The relationship between infection and disease
A systems view of interactions among factors to produce disease
Mechanisms by which disease may alter animal productivity
Effects on ingestion
Effects of disease on feed digestibility
Effects of disease on physiological processes
Measurable effects of diseases on livestock productivity
Effects of disease on herd productivity
Effects on capacity to maintain and improve herd
Effect of disease control measures in productivity of animals
Effects of animal disease on human welfare
Effects of disease on animal welfare
Inclusion of economic effects in a disease model
Conclusion
References


Abstract

For very few disease agents does infection automatically mean that clinical disease will be expressed. In epidemiological terms, for most diseases there are various risk factors which influence whether an animal which is exposed to an agent becomes infected and generates a host response, and a second (frequently overlapping) set of risk factors which determine whether the infection proceeds sooner or later to clinical disease. Risk factors vary widely in their nature, ranging from the genotype of both host and agent, through the nutritional state of the host at the time, to short-term weather conditions at the location where the animals are kept. Traditionally these have been classified into host, agent and environment factors, but I would argue that this is in some respects too static a view of the initiation of disease, and that we should look for a more dynamic way of viewing the interactions. Computer modelling is one way of representing epidemiological interaction realistically and dynamically. While some risk factors are common to many different infectious diseases, others are very specific to a single disease, and it is unwise to extrapolate from knowledge of relevant risk factors for one disease to conclude that the same factors are necessarily important in other superficially similar diseases.

Many diseases exert their effect on productivity most strongly in the early subclinical stages of infection, when there may not necessarily be any evidence of disease, while in others the effects grow as the clinical severity rises. Overall, the scale of effects of disease on productive capacity of animals is surprisingly large, much greater than would occur by depriving the animal of an apparently similar quantity of nutrients. While it is not completely clear why particular diseases have the effect they do on animal productivity, a reasonable overview of why disease so adversely affects productivity has emerged in recent years, and this is presented in this paper. The main effect of disease appears to be on protein metabolism. In addition, some diseases reduce efficiency of utilization of various micronutrients, such as Cu or P. The primary metabolic effects of disease then produce a cascade of secondary effects which can be measured in economic terms. In developing a model of a disease process as a basis for decision-making, consideration must be given to accurately representing these various effects for the particular disease.

The relationship between infection and disease

For very few disease agents does infection automatically mean that clinical disease will be expressed. In epidemiological terms, for most diseases there are various risk factors which influence whether an animal which is exposed to an agent becomes infected and generates a host response, and a second (frequently overlapping) set of risk factors which determine whether the infection proceeds sooner or later to clinical disease. Risk factors vary widely in their nature, ranging from the genotype of both host and agent, through the nutritional state of the host at the time, to short-term weather conditions at the location where the animals are kept.

While some risk factors are common to many different infectious diseases, others are very specific to a single disease, and it is unwise to extrapolate from knowledge of relevant risk factors for one disease to conclude that the same factors are necessarily important in other superficially similar diseases. Risk factors can be identified and the scale of their impact (measured variously as relative risk and attributable risk) assessed through appropriate epidemiological study designs using observational data.

Such studies do not always discriminate between infection and disease, including animals as 'cases of disease' only if they show recognizable clinical signs. However for the types of diseases we are currently seeking to control, it is increasingly necessary to look separately at the process of initial infection, at the process of conversion from infection to disease, and at the ways in which animal productivity is affected at each stage. Similarly for non-infectious diseases, there will tend to be a process of lesion production and a subsequent process of development of clinical disease.

A systems view of interactions among factors to produce disease

Traditionally the influences on disease occurrence have been classified into host, agent and environment factors, but we would argue that this is in some respects too static a view of the initiation of disease, and that we should look for a more dynamic way of viewing the interactions. System diagrams aim to represent through boxes and arrows the pathways by which components of a total biological system interact to create the various processes which drive the system through time, and generate whatever 'outcome' variables are decided by the observer to be of interest. In the case of diseases, the outcome variable may be the prevalence or incidence of the disease as determined by the underlying transmission processes, or it may be the (altered) productivity of the animals when affected by the disease. Through the system diagram approach, not only can host, agent and environment be treated as multifactorial influences in themselves (for example breaking environment into multiple climate variables, shelter, nutrient availability etc.), but the nature of the interactions within and among the various factors can be represented precisely, using arrows only between those factors for which an interaction is thought to occur.

Figure 1 shows one such example of a system diagram developed in order to represent the causal processes and the way in which they are believed to interact to produce a particular disease. The example chosen is one form of lameness in dairy cattle, known as white line disease. This can cause lameness directly but can also lead on to sole abscess which causes even more severe lameness. The figure is based upon research at both herd and individual animal level, and draws together factors which can operate at each of these levels to influence whether or not an individual animal develops a subclinical lesion or becomes clinically lame, and whether the herd has a high or low incidence of clinical lameness. This particular disease is non-infectious, but such representations are even easier to formulate for infectious diseases where the interactions tend to follow more standard patterns.

Figure 1. A path diagram for white line lesions that shows likely causal links between risk factors that predispose to both subclinical and clinical disease.

Such a system diagram is helpful in presenting an understanding of how factors influence disease processes, but it is still static, in that it does not provide any way of analysing and understanding the dynamics of the disease process over time or its spatial spread for diseases which show spatial patterns of occurrence. Computer modelling takes the process a major step further forward and is the most effective way of representing epidemiological interactions realistically and dynamically. Typically nowadays a computer model is designed by first formulating a system diagram, plus sub-diagrams exploring in more detail each specific facet of the disease processes and then programing a computer to mimic the processes and progress the total system forward through time.

In other papers presented at this workshop, techniques of modelling are considered for different types of disease. However if modelling of parasitic diseases is to include representing their economic effects, then it is necessary to formulate an adequate representation of this part of the disease process to complement the epidemiological part of the total model. This paper therefore concentrates on how disease agents affect productivity of animals and how this should be included in a computer model.

Mechanisms by which disease may alter animal productivity

Figure 2 summarizes the various pathways through which disease can adversely affect the productivity of a livestock herd. In the case of infectious and parasitic diseases the underlying principle is that a disease agent is in constant competition with its host for access to nutrient supplies. The agent is successful if it can divert for its own use and reproduction, nutrients which the animal would otherwise have used for growth and production. The agent must therefore have some adverse effects on the host if it is to survive and multiply. Non-infectious diseases cannot be understood in the same simple way, but do frequently represent a change in ecological balance, in which the flow of nutrients and toxins (copper deficiency, facial eczema, etc.) or of controlling signals (hypocalcaemia, ketosis, etc.) through the agricultural ecosystem is distorted by human or environmental interventions of some type. Some of the same principles therefore apply.

The purpose of Figure 2 is to summarize all the possible direct and indirect mechanisms through which a disease can influence the productive efficiency of livestock. Not all diseases will have all of the effects, but in representing economic effects in a model it is necessary to consider all possibilities and select for inclusion those which appear to be relevant. Each of the mechanisms will be discussed individually and then consideration will be given to how they should be combined to evaluate the effect of disease on profitability.

Effects on ingestion

Many diseases alter feed intake in affected animals. In almost all cases intake is reduced (Hawkins and Morris, 1978), but rarely it may be increased (Dargie, 1973). Diseases which cause pain during prehension (contagious ecthyma of sheep) or mechanical difficulty (actinobacillosis of the tongue in cattle) will reduce intake temporarily. Diseases which affect locomotor ability or reduce appetite due to a fever or similar discomfort will also lower intake. However many diseases appear to reduce intake in subtle ways which may not be recognized unless careful measurements are made. These effects have been documented most carefully for parasitic diseases (Larbier et al., 1974), although in some cases intake has been reduced only in more severe forms of the disease (Hawkins and Morris, 1978). Depression of feed intake can also occur in non-infectious diseases such as nutritional deficiencies (Scott et al., 1980).

Figure 2. The various ways in which disease may affect the productive value of animals in a herd or flock.

It is intriguing that feed intake should be commonly depressed by disease when other evidence shows clearly that feed requirements are increased by many of the same diseases, since productivity falls under the influence of the disease. From the limited studies which have been conducted to resolve this apparent paradox, it would appear that it results from disturbances in body homeostatic mechanisms of the host. Symons and Hennessy (1981) have found that choleocytokinin levels rise as appetite falls in Trichostrongylus colubriformis infestations, and return to normal in line with appetite when the infestation is terminated. In the same disease, corticosteroid levels rise and thyroxine levels fall in response to the parasite, while insulin levels fall apparently in response to reduced intake rasher then directly due to the parasite (Prichard et al., 1974; Hennessy and Prichard, 1981). The disease agent may also in some cases produce a toxic substance which depresses intake directly (Seebeck et al., 1971). It is important to differentiate between diseases which merely depress feed intake and those which lower the efficiency of feed conversion - with or without any effect on feed intake. Seebeck et al. (1971) called the effect on intake the anorectic effect and that on feed conversion efficiency the specific effect. The specific effect is the more serious of the two, since lower production is achieved from the same feed intake and efficiency of the production process is adversely affected, whereas the anorectic effect reduces both intake and output without altering the efficiency of production. This differentiation is an important consideration in studies of animals which consume purchased feed, such as pigs. It is less important in grazing ruminants, for which feed production is closer to being a fixed cost.

Effects of disease on feed digestibility

Disease agents do not normally seem to affect feed digestibility, even in the case of diseases which undoubtedly alter the morphology and physiological function of the gastrointestinal tract (Parkins et al., 1973; Reveron et al., 1974). Barker (1974) found that abnormal mucosa was not necessarily linked to poor growth, and it seems that changes in the mucosal surface itself are not responsible for the change in feed conversion efficiency which results from parasitism and other diseases, but rather the physiological processes that occur after absorption. Similar findings have been obtained with parasites such as Fasciola hepatica which do not cause mucosal changes (Hawkins and Morris, 1978). One of the few reports of a reduction in feed digestibility for ruminants was for magnesium deficiency in dairy cows (Wilson, 1980). However the situation may be different in monogastric animals, since two studies of the effects of internal parasites in pigs both showed reductions in feed digestibility (Hale and Stewart, 1979; Hale et al., 1981).

It nevertheless seems likely that, at least in ruminants, adverse effects of disease on productivity which cannot be explained by reduction in feed intake can reasonably be attributed to lower feed conversion efficiency; although as Symons (1969) points out, digestibility trials are a crude method of assessing changes in digestive function. It is also clear that the nature and extent of pathological changes in the body cannot be used as any direct guide to the severity of effects of a disease on productivity.

Effects of disease on physiological processes

Diseases can modify many different physiological processes, such as nutrient metabolism, respiration and excretion. Most of the available data relate to parasitic diseases and the evidence from these studies suggests that the fundamental effect is on protein metabolism.

Steel (1974) and Symons and Steel (1978) have reviewed the metabolic consequences of gastrointestinal parasitism, with particular reference to sheep. They conclude that helminth disease causes a series of metabolic changes to occur in the animal, the primary impact of which is on protein metabolism. The effects, however, carry through to the metabolism of other nutrients. The result is to produce a syndrome analogous to undernutrition.

In gastrointestinal nematode infestations, plasma is lost into the digestive tract at the attachment sites of the parasites, and haemoglobin is also removed by blood-sucking parasites. Much of this protein is digested and reabsorbed lower in the tract, but the host uses energy and protein to replenish the mucosa and plasma proteins which have been depleted. This places demands on the liver and increases its nutrient utilization. There is increased excretion of nitrogen as urea in urine, demonstrating that recycling of the nutrients is not completely efficient in maintaining nitrogen balance, even though considerable energy costs are incurred by the host for increased protein synthesis.

Animals tend under these circumstances to run down their pool of plasma proteins because production in the liver cannot keep pace with the loss, even though the synthesis rate is unusually high. Adjustments are made to other nitrogen-using processes of lower priority, notably synthesis of wool protein and muscle protein. In sheep, sulphur-containing proteins are put in especially short supply by Trichostrongylus colubriformis infestation, demand cannot be met, and wool production shows an exceptionally large fall.

If feed intake is reduced either due to the parasite or to a low plane of nutrition, protein intake may fall below the level required to maintain an adequate serum protein pool. Bown et al. (1986) have shown that direct post-ruminal infusion of casein in sheep receiving daily doses of larvae of Trichostrongylus colubriformis increased nitrogen retention five-fold, and supported the argument as outlined above that the primary defect is one of protein loss and an anabolic cost of tissue regeneration. Infusion of glucose in amounts isocaloric with the casein only doubled nitrogen retention, showing that energy supplementation was not as beneficial as protein replacement.

A contrasting example to Trichostrongylus colubriformis is the cattle tick Boophilus microplus, which sucks blood much like some internal parasites, but differs in that the animal cannot recover any of the nutrient content of the blood in this case. The effects of ticks on host metabolism have been studied by Seebeck et al. (1971), O'Kelly et al. (1971) and Springell et al. (1971). Haemoglobin and plasma albumin fell, whereas globulin rose. Thus the animal was able to synthesize increased supplies of globulins, but could not maintain levels of the other two blood constituents. This was attributed in part to a disturbance of protein metabolism, but the injection of a toxin by the tick was also hypothesized. To further emphasize the tenuous link between the pathology of a disease and its effects on productive processes, O'Kelly and Kennedy (1981) found that ticks adversely affected function in the gastrointestinal tract and reduced organic matter digestibility. It is difficult to explain why this should be so when such effects are not common for parasites directly affecting the tract.

Although these are the two most fully studied diseases, evidence for other diseases in a variety of species confirms the central importance of the derangement of protein metabolism in the disease process. There is also impairment of energy metabolism, but this appears to be largely secondary to the alterations in protein metabolism, and is a result primarily of the energy costs of tissue regeneration.

Mineral and micronutrient metabolic flows are also altered by parasitic diseases, which are the only ones to have been studied. There is reduced retention of ingested calcium and phosphorus in growing sheep infested with Trichostrongylus colubriformis or Ostertagia circumcincta (Symons and Steel, 1978). Consequently, bone growth and skeletal development are impaired, and this can reduce mature body size and capacity to accumulate muscle (Sykes et al., 1977). Cobalt, copper and vitamin status of animals have all been reported to be affected by parasitism (Downey, 1965, 1966a, 1966b) as well.

Since lung disease can adversely affect productivity, another mechanism by which disease might impair physiological function is a reduction in respiratory function. It seems more likely, however, that it is the regenerative process following lung disease which cause the production deficit.

Measurable effects of diseases on livestock productivity

The functional derangements described above translate into measurable economic effects in a number of ways, summarized in Figure 2.

Premature Death

This effect is the easiest of all the consequences of disease to measure, and therefore tends to be considerable over-emphasized in comparison with other effects. In economic studies, death losses should be measured as the difference between the potential market value of the animal and its value when dead (which may not be zero), less the costs which would have been incurred in obtaining the market value (such as extra feed and care to market age, marketing costs, etc.).

Changed Value of Animals and Products From Slaughtered Animals

Diseased animals may have lower market value either due to visible lesions or due to indirect changes in appearance or body conformation which make them less attractive to buyers. True market value of final products may be altered due to changes in the ratio of meat to fat or to bone (Springell et al., 1971; Sykes et al., 1980), or reduced protein content. The value of offals may also be reduced due to pathological changes caused by agents such as Fasciola hepatica or Echinococcus granulosus. Presence of lesions of a zoonotic disease may render the animal totally unfit for consumption.

Some diseases (such as caseous lymphadenitis in sheep) may render products less attractive to the consumer for aesthetic reasons, and hence may reduce meat consumption. Diseases which affect the skin, such as warble fly infestation or even sheep lice (Britt et al., 1986), may reduce the market value of hides or their value to the user.

Reduced Liveweight Gain

There have been well in excess of 50 published studies on the effect of diseases on weight gain in animals and in general they find that diseased animals gain weight more slowly than equivalent disease-free animals. Notable as an exception is lice infestation in cattle. It has been among the most intensively studied but the evidence shows that differences in weight gain between infested and free animals are modest or negligible, and certainly not enough to yield an economic benefit from treatment. Therefore caution is required in assuming an effect on weight gain of a disease without experimental data to support it.

Reduced Yield and Quality of Products From Live Animals

Yield of products such as milk, wool and eggs may also be reduced by disease, and there have been numerous papers showing the effect of various diseases on wool growth or milk-yield. Quality of the products may also be reduced, as in the case of the changes in milk composition which result from bovine mastitis, and these may or may not be detectable by the consumer. In the first case price will fall and the livestock producer will suffer; in the second case, the consumer will suffer the loss. For example, parasitic disease can reduce the market value of wool per kg, as well as the quantity produced (Morris et al., 1977); but the structural characteristics of the wool may also be altered in ways which reduce its value to the manufacturer but cannot be detected in the normal marketing process (Johnstone et al., 1976). It has also been shown that parasitic disease can affect the taste of meat (Garriz et al., 1987).

Reduced Capacity for Work

Worldwide, the single most important use of animals is as a source of traction. The second largest (after dung) productive energy output of animals in developing countries is for work, and products considered as of central importance in developed countries are seen as bi-products under those conditions (Odend'hal, 1972). There have been no published reports directly measuring the effects of diseases on capacity for work, but field evidence is that diseases can severely curtail rice paddy preparation and other tasks for which animals are essential, so this effect can be very important and should be considered in developing countries.

Altered Production of Dung for Fuel and Fertilizer

In Asia and Africa cattle dung is a vital source of cooking fuel and in much of the developing world it is an important fertilizer. Diseases which cause high death rates in cattle will also indirectly influence human nutrition by reducing dung supplies.

Altered Feed Conversion Efficiency

As discussed earlier, it appears that disease primarily affects animal productivity by altering the metabolic processes for protein and other nutrients, thereby reducing the feed conversion efficiency of affected animals and producing a number of ramifications which reduce herd productivity. Feed intake may also be reduced, but this is not usually the primary effect.

Feed conversion efficiency is the ultimate measure of the influence of disease on the production process, but its measurement requires accurate measurement of feed intake, and that is only possible under controlled feeding conditions. In grazing systems it is usually reasonable to take changes in productivity as an adequate indication of changes in feed conversion efficiency when comparing diseased and disease-free animals kept under identical conditions.

Intuitively, it seems likely that the rate of decline in productivity would increase as the disease becomes more severe and body functions become more deranged. However, the limited evidence available favours the alternative view that the most dramatic changes occur at low or subclinical levels of disease, and that each additional parasite, for example, has less effect than the one before it (Hawkins and Morris, 1978). This emphasizes the importance of the health management approach in which the focus is on optimizing productive efficiency rather than the clinical approach in which a disease must be detectable to be considered important.

Effects of disease on herd productivity

The effects of disease flow through from consequences for individual animals to broader ramifications for herd replacement and improvement.

Reduced Productive Life of Animals

Apart from animals which die, all remaining herd members are culled when the manager considers them less potentially productive than the animal which would replace them. This issue has been investigated in detail by Renkema and his co-workers (Renkema and Stelwagen, 1979; Korver and Renkema, 1979; Dijkhuizen et al., 1985a, 1985b). They showed that in general a substantial economic benefit could be achieved by taking action to extend the herd life of the average dairy cow, principally by reducing the amount of involuntary culling due to health-related causes. This is not limited to disposal specifically because of disease, but also includes culling for low yield or other causes, where the underlying cause is lowered productivity due to disease, but the manager is unaware of this fact.

Less Accurate Genetic Selection

If a disease alters any of the components of productivity which are the subject of genetic selection pressure in the herd (such as milk or wool yield), it will affect the efficiency with which animals of superior genetic merit are identified, especially if the probability of an animal being affected by the disease is unrelated to yield level. Provided susceptibility to the disease and yield level are not correlated, the presence of the disease will confound the genetic selection effort. For example, Johnstone et al. (1976) showed that internal parasitism can affect wool production by sheep in ways which distort selection by objective measurement of wool characteristics. Since resistance to internal parasitism cannot be regarded as a heritable trait for practical purposes, genetic selection will be more efficient if effective parasite control is being carried out in the herd.

Effects on capacity to maintain and improve herd

If fewer progeny are born, less animals are available as herd replacements or for sale as market products. Thus not only will livestock sale income be reduced, but management flexibility for herd improvement will be curtailed. It is self-evident that diseases of the reproductive tract in both males and females can substantially reduce the level of reproductive performance, and hence the number of progeny born in the herd.

Less obviously, diseases which adversely affect body metabolism (but do not directly affect the reproductive tract) can also affect the number of progeny born. The mechanisms have not been fully explored, but may well operate through an effect on liveweight and condition, or through indirect means such as the induction of pyrexia at critical stages in the reproductive process. For example, both gastrointestinal parasites (Murray et al., 1971) and liver fluke (Hope Cawdery, 1976) have been shown to affect reproductive performance in ewes. In cattle, bovine leucosis (Schmied et al., 1979; Parchinski, 1979) and ephemeral fever (Theodoris et al., 1973) have been reported to affect reproduction. If reproductive performance is too poor, it may even become impossible to maintain herd size through home-bred replacements, necessitating the purchase of breeding animals with all the additional risks which that entails.

Effect of disease control measures in productivity of animals

In evaluating the economic benefit of disease control, it is necessary to consider not only the difference in productivity between diseased and disease-free animals, but also the changes in productivity which follow elimination of a disease from an affected animal.

This has not been studied for very many diseases, but some examples exist. For instance, bovine mastitis appears to be a disease for which complete regeneration occurs in most animals over the dry period following elimination of an infection (Morris, 1973), although yield remains depressed for the rest of the lactation in which a cure is achieved. Conversely, when infestations with Fasciola hepatica are eliminated in growing animals, sheep do not regain their former productivity or feed conversion efficiency, even when the infestation had existed for as little as eight weeks (Hawkins and Morris, 1978). In a study of a nematode parasite, wool growth and liveweight gain responded quite differently to anthelmintic treatment (Coop et al., 1984).

Therefore each disease type must at least in the first instance be considered separately, since the nature and extent of recovery following elimination of a disease is not predictable from general principles. The selection of an economically optimal control strategy will be strongly influenced by this consideration.

Effects of animal disease on human welfare

Effects on Human Nutrition

The major direct effect of animal disease on human well-being is through reducing the supply of high quality protein, such as diseases which reduce the supply of milk for young children. Animal products are also important sources of other nutrients, notably minerals and vitamins, and diseases can both reduce the total supply of animal products and modify the composition of animal products in ways which reduce their nutritional value (Huss-Ashmore and Curry, 1992).

Effects on Community Development

As well as the effects on human nutrition, animal diseases can affect other aspects of community welfare, especially in developing countries. As discussed earlier, the two most important services provided by animals in such circumstances are traction and dung production, and disease may reduce the supply of both of these. Animals are also important sources of products (wool, hair, hides, feathers, fur, etc.) used for clothing, decoration and for manufacture of utensils and other products. A further effect of those animal diseases which are zoonotic is to cause disease in the human as well as the animal population, thus amplifying their impact.

Cultural Significance of Animals

In most communities animals serve functions far beyond the utilitarian roles which are the focus of this paper. While these are not strictly economic in nature, they are vital functions which should be included in any consideration of the significance of animal disease.

Effects of disease on animal welfare

In considerations of animal welfare issues, little is said about the importance of ensuring through disease control that animals are in a healthy state - yet this is a vitally important issue in protecting the welfare of managed animals. It deserves more prominent attention in discussions of animal welfare matters.

Inclusion of economic effects in a disease model

The formulation of an economic analysis for parasitic diseases is described by Morris and Meek (1980). Meek and Morris (1981) showed how economic components could be built into a computer model, which could then be used to conduct comprehensive evaluations of alternative parasite control programs.

Provided that the relevant items from Figure 2 are included into the total simulation model by linking them to the appropriate ecological and epidemiological indices within the overall system model, then estimation of the benefits of disease control strategies within the simulation model becomes straightforward, and depends only on the availability of suitable field data on the effects of the disease on various yield measures. If such data are unavailable when the model is formulated, guesstimates can be used initially and sensitivity analysis applied to determine which of the productivity indicators most urgently need field refinement.

Estimation of costs at farm level can usually be done quite simply from information on the unit costs of control measures, since a partial budgeting approach to such economic analyses is almost always adopted at this level.

In regional evaluations it may be necessary to build a complete model of a regional disease control program. However in many cases it is sufficient to run the model for various types of farms and sets of conditions and then to combine these through an electronic spreadsheet in which the regional total effects are calculated, taking into account costs above the farm level and possible supply/demand consequences of disease control programs.

Conclusion

In building a computer model of a parasite control issue, economic components can readily be included in the total model formulation, provided that their inclusion is thought through from the start and conforms to current understanding of the ways in which disease influences productivity. In this way a computer model can examine not only the epidemiological consequences of a disease, but also the economic effects at farm and regional or national level which might flow from possible control programs.

References

BARKER, I.K. 1974. Relationship of abnormal mucosal microtopography with distribution of Trichostrongylus colubriformis in the small intestines of lambs. International Journal of Parasitology 4: 153-163.

BOWN, M.D., POPPI, D.P. and SYKES, A.R. 1986. The effect of post-ruminal infusion of protein or energy on the pathology of Trichostrongylus colubriformis infection on body composition in lambs. Proceedings of the New Zealand Society of Animal Production 46: 27-30.

BRITT, A.G., COTTON, C.L., PITMAN, I.H. and SINCLAIR, A.M. 1986. Effects of the sheep-chewing louse (Damalinia ovis) on the epidermis of the Australian Merino. Australian Journal of Biological Sciences 39: 137-143.

COOP, R.L., ANGUS, K.W., HUTCHINSON, G. and WRIGHT, S. 1984. Effect of anthelmintic treatment on the productivity of lambs infected with the intestinal nematode Trichostrongylus colubriformis. Research in Veterinary Science 36: 71-75.

DARGIE, J.D. 1973. Ovine haemonchosis: pathogenesis. In: Urquhart, G.M. and Armour, J., eds. Helminth Diseases of Cattle, Sheep and Horses in Europe. Glasgow: University Press, pp. 63-72.

DIJKHUIZEN, A.A., RENKEMA, J.A. and STELWAGEN, J. 1985a. Economic aspects of reproductive failure in dairy cattle. I. Financial loss at farm level. Preventive Veterinary Medicine 3: 251-263.

DIJKHUIZEN, A.A., RENKEMA, J.A. and STELWAGEN, J.U. 1985b. Economic aspects of reproductive failure in dairy cattle. II. The decision to replace animals. Preventive Veterinary Medicine 3: 265-276.

DOWNEY, N.E. 1965. Some relationships between trichostrongylid infestation and cobalt status in lambs. I. Haemonchus contortus infestation. British Veterinary Journal 121: 362-370.

DOWNEY, N.E. 1966a. Some relationships between trichostrongylid infestation and cobalt status in lambs. II. Trichostrongylus axei infestation. British Veterinary Journal 122: 201-208.

DOWNEY, N.E. 1966b. Some relationships between trichostrongylid infestation and cobalt status in lambs. III. Trichostrongylus axed and Ostertagia circumcincta infestation. British Veterinary Journal 122: 316-324.

GARRIZ, C.A., GALLINGER, M.M., TOURAILLE, C. STEFFAN, P.E., FIEL, C.A., AMBRUSTULO, R.R., BIONDANI, C.A., ZAMORANO, M. and BULMAN, G.M. 1987. Gastrointestinal parasitism: its effects on muscle, fat and bone composition of the carcass and organoleptic characteristics of meat. In: Proceedings MSD Agvet Symposium on Gastrointestinal Parasitism, August 19, 1987, pp. 59-68.

HALE, O.M. and STEWART, T.B. 1979. Influence of an experimental infection of Trichuris suds on performance of pigs. Journal of Animal Science 49: 1000-1005.

HALE, O.M., STEWART, T.B., MARTI, O.G., WHEAT, B.E. and McCORMICK, W.C. 1981. Influence of an experimental infection of nodular worms (Oesophagostomum spp.) on performance of pigs. Journal of Animal Science 52: 316-322.

HAWKINS, C.D. and MORRIS, R.S. 1978. Depression of productivity in sheep infected with Fasciola hepatica. Veterinary Parasitology 4: 341-357.

HENNESSY, D.R. and PRICHARD, R.K. 1981. Functioning of the thyroid gland in sheep infected with Trichostrongylus colubriformis. Research in Veterinary Science 30: 87-92.

HOPE CAWDERY, M.J. 1976. The effects of fascioliasis on ewe fertility. British Veterinary Journal 132: 568-575

HUSS-ASHMORE, R. and CURRY, J.J. 1992. Impact of improved livestock disease control on household diet and welfare: a study in Uasin Gishu District, Kenya. In: ILRAD Technical Report No. 2. Nairobi: International Laboratory for Research on Animal Diseases, 97 pp.

JOHNSTONE, I.L., DARVILL, F.M. and SMART, K.E. 1976. The influence of parasites on selection parameters in sheep. In: Tomes, G.J., Robertson, D.E. and Lightfoot, R.J., eds. Proceedings of 1976 International Sheep Breeding Congress. Perth: West Australian Institute of Technology, pp. 256-262.

KORVER, S. and RENKEMA, J.A. 1979. Economic evaluation of replacement rates in dairy herds. II. Selection of cows during the first lactation. Livestock Production Science 6: 29-37.

LARBIER, M., YVORE, P. and GUILLAUME, J. 1974. Influence of duodenal coccidiosis on the utilization of dietary energy and protein in chickens. Annals de Recherches Veterinaires 5: 179-188.

MEEK, A.H. and MORRIS, R.S. 1981. A computer simulation model of ovine fascioliasis. Agricultural Systems 7: 49-77.

MORRIS, R.S. 1973. The depression of quarter milk yield caused by bovine mastitis, and the response of yield to successful therapy. Australian Veterinary Journal 49: 153-156.

MORRIS, R.S. and MEEK, A.H. 1980. Measurement and evaluation of the economic effects of parasitic disease. Veterinary Parasitology 6: 165-184.

MORRIS, R.S., ANDERSON, N. and McTAGGART, I.K. 1977. An economic analysis of two schemes for the control of helminthiasis in breeding ewes. Veterinary Parasitology 3: 349-363.

MURRAY, J., LEANING, W.H.D and MARTIN, C.A. 1971. Pre-mating anthelmintic treatment of ewes and its effects on lambing performance. New Zealand Veterinary Journal 19: 1-4.

O'KELLY, J.C. and KENNEDY, P.M. 1981. Metabolic changes in cattle due to the specific effect of the tick Boophilus microplus. British Journal of Nutrition 45: 557-566.

O'KELLY, J.C., SEEBECK, R.M. and SPRINGELL, P.H. 1971. Alterations in host metabolism by the specific and anorectic effects of the cattle tick (Boophilus microplus). II. Changes in blood composition. Australian Journal of Biological Sciences 24: 381-389.

ODEND'HAL, S. 1972. Energetics of Indian cattle in their environment. Human Ecology 1: 3-22.

PARCHINSKI, O. 1979. Effects of bovine leukosis on reproductive function in cows. Trudy Latviiskoi Sel'skokhoozyaistrennoi Akademii 169: 40-47.

PARKINS, J.J., HOLMES, P.H. and BREMNER, K.C. 1973. The pathophysiology of ovine ostertagiasis: some nitrogen balance and digestibility studies. Research in Veterinary Science 14: 21-28.

PRICHARD, R.K., HENNESSY, D.R. and GRIFFITHS, D.A. 1974. Endocrine responses of sheep to infection with Trichostrongylus colubriformis. Research in Veterinary Science 17: 182-187.

RENKEMA, J.A. and STELWAGEN, J. 1979. Economic evaluation of replacement rates in dairy herds. I. Reduction of replacement rates through improved health. Livestock Production Science 6: 15-27.

REVERON, A.E., TOPPS, J.H., MacDONALD, D.C. and PRATT, G. 1974. The intake, digestion and utilization of food and growth rate of lambs affected by Trichostrongyulus colubriformis. Research in Veterinary Science 24: 947-953.

SCHMIED, L.M., PAULI, R., RIBET, A., RIBER, S.S. and ALOISI, G. 1979. First serological confirmation of bovine leukosis in Argentina. Its effects on the fertility of cows. Revista de Medicina Veterinaria, Argentina 60: 72-80.

SCOTT, P.R., KELLY, J.M., WHITAKER, D.A. and CAMERON, N.D. 1980. Marginal magnesium deficiency as a possible cause of reduced voluntary intake in commercially managed dairy cows. Veterinary Research Communications 4: 225-229.

SEEBECK, R.M., SPRINGELL, P.H. and O'KELLY, J.C. 1971. Alterations in host metabolism by the specific and anorectic effects of the cattle tick (Boophilus microplus). I. Food intake and body weight growth. Australian Journal of Biological Sciences 24: 373-380.

SPRINGELL, P.H., O'KELLY, J.C. and SEEBECK, R.M. 1971. Alterations in host metabolism by the specific and anorectic effects of the cattle tick (Boophilus microplus). III. Metabolic implications of blood volume, body water and carcass composition changes. Australian Journal of Biological Sciences 24: 1033-1045.

STEEL, J.W. 1974. Pathophysiology of gastrointestinal nematode infections in the ruminant. Proceedings of the Australian Society of Animal Production 10: 139-147.

SYKES, A.R., COOP, R.L. and ANGUS, K.W. 1977. The influence of chronic Ostertagia circumcincta infection on the skeleton of growing sheep. Journal of Comparative Pathology 87: 521-529.

SYKES, A.R., COOP, R.L. and RUSHTON, B. 1980. Chronic subclinical fascioliasis in sheep: effects on food intake, food utilization and blood constituents. Research in Veterinary Science 28: 63-70.

SYMONS, LEA and HENNESSY, D.R. 1981. Cholecystikin and anorexia in sheep infected by the intestinal nematode Trichostrongylus colubriformis. International Journal of Parasitology 11: 55-58.

SYMONS, LEA and STEEL, J.W. 1978. Pathogenesis of the loss of production in gastrointestinal parasitism. In: Donald, A.D., Southcott, W.H. and Dineen, J.K., eds. The Epidemiology and Control of Gastrointestinal Parasites of Sheep in Australia. Melbourne: CSIRO, pp. 9-22.

SYMONS, LEA. 1969. Pathology of gastrointestinal helminthiases. International Review of Tropical Medicine 3: 49-100.

THEODORIS, A., GIESECKE, W.H. and du TOIT, I.J. 1973. Effects of ephemeral fever on milk production and reproduction of dairy cattle. Onderstepoort Journal of Veterinary Research 40: 83-92.

WILSON, G.F. 1980. Effects of magnesium supplements on the digestion of forages and milk production of cows with hypomagnesaemia. Animal Production 31: 153-157.

Modelling livestock productivity

J. McIntire

The World Bank
1818 H Street, NW
Washington DC 20433, USA

Abstract
Pasture production
Model prices
The economic effects of diseases
The effects of risk
Responses to risk
Offtake rules
Initial model values
Results
Prices determined by world market
Prices determined by domestic market
Discussion
Efficiency of offtake rules
Incentive problems
Future research
References


Abstract

What are the main economic issues in modelling cattle disease control? Many diseases reduce the productivity of African livestock production. Among those diseases are several borne by such vectors as ticks and biting flies. While much is known about the biology and control of such diseases, and about the vectors that transmit them, little is known about related economic problems. Economic issues include the expected returns to disease control, production risks and their relations to the returns to disease control, and the incentive problems involved in disease control programs.

This paper examines those economic issues in the following manner. First, an economic model of livestock production is presented which allows analysis of the effects of stock diseases. Second, results from the model are presented under different market structures and levels of disease. Third, attention is given to risk and incentive problems faced by producers which might make them unwilling to adopt apparently promising disease control methods. The fourth and final section is a summary and conclusion.

MODELLING CATTLE PRODUCTION

This section sketches a model of cattle production under conditions representative of the semiarid rangelands of Africa, extending that developed by von Kaufmann et al. (1990). The notation is:

Variables:

q = number of animals in the herd
Z = quantity of non-veterinary variable inputs
V = quantity of veterinary inputs
X = long-term average quantity of pasture production
M = long-term average consumption of pasture by cattle
S = a stock of L, Z, V, or X
m = annual consumption of pasture by livestock
x = annual quantity of pasture production
b = number of animals sold annually (physical offtake)
q = percentage of herd sold annually (offtake rate)
d = economic losses to animal disease
p = profit or net revenue

Parameters:

f = rate of natural herd growth
p = price of liveweight
c = market cost of non-veterinary inputs
c1 = market cost of fixed veterinary inputs
c2 = market cost of variable veterinary inputs
u = transport costs from domestic to international market
n = an elasticity
F = average annual intake of pasture per tropical livestock unit*
a = probability that vector will transmit pathogen to herd
b = probability that an animal will become infected
G = morbidity factor (response of animal productivity to disease)
r = real rate of discount

* A tropical livestock unit (TLU) is the equivalent of 250 kg animal.

Indices:

t = year
i = animal class (e.g., four-year-old males)

Suppressing the subscript 'i' for animal class, the general form of the herd's growth is:

(1) qt+1 = qt* (1 + f t - q t)

By assumption producers can purchase no stock+, so q ³ 0, and, by definition, they cannot sell more than they own, so q £ 1. The parameter f is obviously ³ 0, meaning that herd growth has to be non-negative. The profit (or net revenue) function++ corresponding to equation (1) is:

(2) p t = Ptq tqt - cZt.

Because animals can be sold now or in the future, the income from their sale depends not only on the offtake in the current year, but in all future years as well. This necessitates modifications in equation (2) to maximize the present value of net revenue (NPV) over all years, as follows

(3) NPVp = S t (Ptq tqt - cZt)*e-rt, where 'r' is the real rate of discount.

* A tropical livestock unit (TLU) is the equivalent of 250 kg animal.

+ The justification for this assumption is that uninsurable risks (sometimes called moral hazard) in the market for breeding stock make it too risky to buy animals for anything other than immediate consumption (Binswanger and McIntire, 1987).

++ Milk production is excluded.

Pasture production

Pasture and other roughages, such as crop residues and browse, are the principal feed of most African livestock. The model treats pasture as a random variable because feed supply is typically not under the control of the producer. The annual quantity of pasture, where the latter refers to the total of all available roughages, is given by

(4a) xt = X + m t,

where m t is the random deviation (mean of 0) from long-term average pasture (X) availability.

The annual consumption of pasture by livestock is

(4b) mt = qt*F.

In long-run equilibrium, all pasture is consumed by stock (i.e., M = X). Each year producers have to apply rules of offtake from their herds so as to arrive at a herd size feasible with available pasture.* Two offtake rules are discussed in a later section.

* A recent paper allows for feedback between qt and xt in which qt > Q causes xt+1 < xt. This is not done in the present model.

Model prices

Prices of liveweight can be determined in two different market structures. One structure is to allow domestic prices to be determined by world prices and transport costs, as in

(5a) pt = P + u,

where the price in year 't' is always equal to the long-term average world market price (P) plus transport costs (u) from world to domestic markets. This is called the world market structure.+

+ This specification assumes that the country is a net importer of beef; if it were a net exporter, then the equation would be pt = P - u.

An alternative structure is to allow domestic prices to be determined generally by herd offtake within certain limits, as in

(5b) pt = f(b),

where dp/db < 0.

The values of pt are bounded within limits set by producers' behaviour and transport costs. Initially,

(5c) pt = (bt-1)e-n,

where bt-1 = qt-1 *q t-1, qt-1 is the average quantity produced in the previous year, q t-1 is the offtake rate in the previous year, n (> 0) is the price elasticity of demand for beef, and pt is the market price in year 't'. The values of pt from equation (5c) are constrained by:

(5d) minimum of pt = P - u, and

(5e) maximum of pt = P + u.

Equations (5c) through (5e) say that the domestic market price is determined solely by offtake in the previous year. It is independent of world prices, unless the domestic price falls below P - u, or if it rises above P + u.

Market prices of inputs (the parameters p, w, c, and r) are assumed to be unaffected by herders' actions and to be constant in real terms over all years.

The economic effects of diseases

The economic effects of animal disease on livestock productivity depend, in general, on the probability of disease transmission and subsequent infection, the effects of infection on the productivity of individual animals, and the costs and benefits of control measures available to producers. The notation for the disease process is:

t = the economic loss associated with vector borne disease;
a = the probability of disease transmission to any individual in herd by a vector;
b = the probability of infection of an individual animal in a herd;
G = the morbidity of infected animals.

A general representation of the economic loss is:

(6) t = f(G,V,p,c1,c2).

In equation (6) the economic loss depends on the physical costs of morbidity in infected animals (parameter G), the quantity of veterinary inputs (V), the price of livestock output (p), and the costs of veterinary inputs (fixed, c1, and variable, c2). The specification is

(7) t = p(G (V)) - c1 - c2V

The loss to the producer caused by disease is the value of productivity losses P(G (V)) minus the fixed and variable costs of veterinary inputs. Morbidity, G, is a function of the probabilities of transmission and of infection, the maximum morbidity (K) relative to the value of the animal, the quantity of variable disease control inputs (V), and the effect of such inputs on morbidity (the parameter 'n') as follows

(7a) G = a b K(1 - V-n)

Substituting into (7) with (7a) gives

(7b) t = pa b K(1 - V-n) - c1 - c2V.

Given that K º 1, that a and b are bounded by 0 and 1, and V ³ 0 implies that G is also bounded by 0 and 1. Dividing (7b) by p converts t into the relative loss of output caused by livestock disease, as in

(8) t ' = a b K(I - V-n) - c1/p - (c2/p)V

If there are no fixed or variable costs (i.e. V = c1 = 0), then t ' = a b K.

Accordingly, if it were certain that the disease would be transmitted and cause infection (i.e. a = b = 1), then t ' would also be equal to 1.

Substituting t ' into (3) gives

(9) NPVp = S t [pqt (q t - t 't) - CZt]*e-rt.

The probability of transmission, a, varies among herds as some are exposed to areas infested with vectors (for example tick-infested pastures) while others are not exposed. It is expected to be positively correlated with the numbers of animals per unit of land, being low on land with few animals and high on lands with many animals.

The probability of infection, b, does not vary among the individuals in a herd because the animals are assumed to have similar genetic makeups. The random variations between individuals in the herd are further assumed to be unrelated to producers' decisions. The probability that an animal will become infected is then simply the product of the probability that the vector will transmit the disease to an individual times the probability that the affected individual will become infected.

The calculation of equation (9) is presented in Table 2. First, values of a and b are postulated. The parameter 'K' is set to 1 to indicate the maximum loss from morbidity. The marginal effect of variable inputs on morbidity from disease ('n') is specified at both high and low values. The optimal level of veterinary inputs is then calculated from the morbidity function, and profit at that optimal level is calculated from equation (9).

The effects of risk

African livestock producers obviously face important risks to their herds from many factors. Risks considered here are price, pasture production and disease.

Most producers will have some degree of aversion to risk. The cost of such risk aversion can be understood as the loss of utility caused by the natural variation in revenue from a productive activity; although some producers may prefer such variation, they are not considered here. The calculation of this utility loss is done by modifying the profit function for the herd (equation 3) to express it in terms of utility. The relation between utility and profit (or some measure of income or consumption) is known as a utility function. One common functional form is the logarithmic (Anderson et al., 1977),* as in

(10) U = loge (W0 + p) - 0.5p 2/(W0 + p)2 + 0.33p 3/(W0 + p)3

where p is the mean, p 2 the variance, and p 3 the skewness of the net present value of profits derived from equation 9. The variable W0 is the initial wealth of the herd owner, defined as the total value of livestock at the beginning of the model period. Utility (U) in this formulation is always affected positively by mean profit and initial wealth, and always negatively by increasing variance of profits (because p 2 ³ 0 by definition).

* There are many others. The logarithmic utility function of wealth has the desirable property of lower risk aversion with higher wealth.

Table 1. Initial parameter values used in the model.

Class

Numbers

Weight(kg)

Inputs


Labour input (L), years/animal

.10



Non-veterinary inputs (Z), units

1



Veterinary inputs (V), units

1



Number of animals

25


Costs


Wage (w), $/year

20



Variable cost of veterinary inputs (c2),

$.50



Prices




Price of livestock output (p), $/mt

$850



Discount rate (r)

10%


Other parameters


Herd growth rate (f)

20%



Elasticity of livestock product demand (p)

-.70


Herd structure and stock weights


Cows

9

186


Calves, 0-1 years

3

62


Calves, 1-2 years

3

125


3 year females

2

150


4 year females

2

186


3 year males

2

175


4 year males

2

190


5 year males

1

190


6 year males

1

190


7 year males

0

190


8 year males

0

190


9 year males

0

190


10 year males

0

190

The effect of skewness in profits on utility is ambiguous. The effect is positive (negative) if the skewness is itself positive (negative). For example, the distribution of profits is positively skewed if the herd owner experiences a few very good years and many years a little below the average. It is even conceivable that a producer in a highly skewed environment would accept negative mean profits if the skew of profits were great enough.

Table 2. Calculation of loss to morbidity.



Disease Pressure

High

Low

Risky

Probability of transmission

.90

.30

.67

Probability of infection

.90

.30

.67

Maximum (K)

1.00

1.00

1.00

Fixed cost (c1)

.00

.50

.00

Variable cost (c2)

.50

.10

.50

Marginal physical product of variable input (n)

-1.00

-2.00

-1.00

Price (P)

.70

.80

.90

Profit @ 0 input use





absolute

-.57

-.57

-.41


relative to price

-.81

-.72

-.45

Optimal input use

1.06

1.13

.90

Morbidity @ optimal input

.76

.07

.50

1 - morbidity @ optimal input

.24

.93

.50

Profit @ optimal input use


absolute

-.40

-.55

-.25


relative to price

-.57

-.68

-.28

Responses to risk

African herders have few responses to the risks of pasture production. Because it is generally uneconomic to produce forage crops or other substitutes for risky pasture production (McIntire et al., 1992), risk avoidance most often includes diversification of species or breeds, stock mobility, and varying offtake as a function of available pasture. The only such strategy analysed here is varying offtake, as noted in the following discussion of offtake rules.

Herders have even fewer alternatives for avoiding the risks caused by livestock disease. Herd diversification might not be effective, as different breeds might often be susceptible in similar degree to the same disease. Mobility may only be a seasonal alternative for avoiding disease and can be limited by availability of pasture. The sole strategy for reducing the effects of disease and thereby cutting risks of livestock morbidity and death is to increase the use of veterinary inputs.

Offtake rules

Herd offtake changes as a function of price and climatic variations that affect pasture productivity and subsequently herd growth. One way to model offtake changes is to solve equation (3) by setting constraints to the parameter q. Those constraints are known as offtake rules (McIntire, 1991), of which two are analysed here.

Partial Feed Equilibrium Rule

With this rule, the manager adapts offtake to the average availability of pasture. Owners sell no animals if pasture production is above the long-term average; but sell some when pasture production is below the long-term average. This means either

(11a) if xt > X, bt = 0; or
(11b) if xt £ X, bt = - (xt - X)/F,

where bt is the number of animals sold in year 't'. The rationale for this rule is the observation that producers build up their herds during good years when pasture production is above average, and that they sell animals when pasture production is below average because they expect that the animals will lose weight or die for lack of pasture. Admittedly, the pasture produced above the long-term average ([xt > X]) is wasted in this specification.

Target Breeding Herd Rule

With this offtake rule, the herd manager sells a number of breeding females sufficient to maintain an initial number at the optimal expected sale age. This is a formulation of the argument of Dahl and Hjort (1976) that producers seek to minimize the risk of falling below a minimum herd size. The target breeding herd rule means

(11c) bjt = qjt - qj0

where the subscript 'j' refers to breeding females and the term qj0 is the number of breeding females in the herd at the beginning of the model period. The term qj0 is set proportionately equal to values observed in African field studies (Itty, 1992). The offtake rules in equations (11a) and (11b) apply to all other herd classes.

Initial model values

In summary, the model has three state variables - price, pasture production and morbidity - and two control or decision variables - offtake and the quantity of veterinary inputs used. The model can be run with deterministic or stochastic values of the state variables. The means of price and pasture production are used for their deterministic values. Somewhat hypothetical values of the morbidity variables were used in the deterministic version. For the stochastic value of pasture output, a random number generator is used to simulate a positively skewed distribution. The resulting values of xt subsequently determine pt the other state variable, through equations (5c), (5d) and (5e). A sample size of 50 is used in the stochastic simulations.

Representative initial values are shown in Table 1. The livestock breed is assumed to be zebu cattle. Cattle are herded on ranges and receive little supplementary feed, veterinary care, or non-feed inputs except labour. The initial herd size is 25 adult animals.

The cost of non-veterinary inputs is US$ 0.50/unit, based on research in the ATLN (Itty, 1992). The world price of one metric tonne of liveweight is US$ 850, or US$ 213 for an animal of 250 kg liveweight (Itty, 1992).

Results

The model was used in several experiments to test the effects of the parameters, m, b, n, P, and c1 on producers' demands for veterinary inputs under different scenarios about offtake rules. The calculation of the morbidity function and the related profit function for disease control is given in Table 2. Scenarios and some initial results are in Table 3.

Prices determined by world market

In the unlikely case of no animal disease, fixed pasture production, and fixed output prices, the only decision variables are the offtake rules because disease control is unneeded by definition. Income in scenario 1 would be US$ 4,300 with the target offtake rule and US$ 4,700 with the feed equilibrium rule.

If pasture production is variable, then incomes are lower still, under both offtake rules. The relative variability of income is much higher under the feed equilibrium rule, both in relation to the offtake rule in this scenario and with respect to the feed equilibrium rules in the preceding scenario.

In scenario 2, pasture production is variable, but world prices still hold, and there is no animal disease. Incomes are much lower than in scenario 1 because the necessity of adjusting offtake (with either rule) to feed availability means that animals cannot always be sold at the optimum age, as they could be if pasture production were constant, or if there were sources of supplementary feed. Income is slightly lower and slightly more variable with the feed equilibrium rule than with the target rule, and the latter rule has a higher probability of negative income.

Despite a low level of stock disease in scenario 3, mean income, relative income variability and utility do not change from scenario 2 except for some minor variation due to sampling error in the simulations. In scenario 4 (high disease), mean income and utility fall, while the relative variability of income rises. Minimum income, a measure of the worst risk to the producer, falls in scenario 4.

Prices determined by domestic market

A more realistic group of scenarios is one in which domestic prices diverge from world prices. Even in the absence of animal disease, income will fall. The results (scenario 5 in Table 3) show that income would be much lower than in scenario 2 and, in particular for the feed equilibrium offtake rule, that the relative variation of income is greater.

Table 3. Offtake rules (NPV of income or utility).

Scenario

Target

Feed Equilibrium

1. Assured pasture production, world prices no disease


Mean

4.30

4.70


Utility

4.77

5.20

2. Variable pasture production, world prices, no disease


Mean

2.88

2.63


Minimum

-.33

-.39


Cv

30.4%

35.7%


Utility

3.67

3.52

3. Variable pasture production, world prices, low disease


Mean

2.91

2.67


Minimum

-.32

-.38


Cv




Utility

3.71

3.56

4. Variable pasture production, world prices, high disease


Mean

2.07

1.74


Minimum

-.47

-.60


Cv

33.6%

45.8%


Utility

2.77

2.58

5. Variable pasture production, endogenous prices, no disease


Mean

2.13

1.79


Minimum

-.31

-.55


Cv

31.2%

43.4%


Utility

2.81

2.61

6. Variable pasture production, endogenous prices, low disease


Mean

1.94

1.59


Minimum

-.30

-.58


Cv

31.5%

46.6%


Utility

2.60

2.40

7. Variable pasture production, endogenous prices, high disease


mean

1.36

.96


Minimum

-.45

-.73


Cv

36.7%

67.9%


Utility

2.05

1.87

8. Variable pasture production, endogenous prices, variable disease


Mean

1.94

1.58


Minimum

-.27

-.55


Cv

31.5%

46.9%


Utility

2.60

2.40

Income is lowest and most variable under high disease pressure (scenario 5). Income is lowest and most variable with the interaction of variable pasture production, endogenous prices, and high disease pressure (scenario 7). Partitioning the 68% fall in income from the best scenario (1) to the worst (7), shows that 33 % is due to moving from assured pasture production to variable (scenario 1 to 2), another 17% is due to moving from world to endogenous prices (scenario 2 to 5), and another 18% is due to moving from no disease to high disease (Scenario 5 to 8).

A likely scenario is one in which disease is random, unknown to the producer in advance. Herd managers will have expectations about the likelihood of disease, but cannot know its exact level. Therefore, they must make decisions about veterinary inputs without full knowledge about the likelihood of disease. They may also lack complete knowledge of the effects of veterinary inputs on stock productivity.

The scenario in which disease is random is as follows. To simplify the exercise, the probability of infection is assumed to represent both the probability of disease transmission to a member of a herd and the subsequent infection of the animal. Equations (8a) and (9) can then be used to derive the expected value of V, the level of variable inputs, given that b is random.

(12) E (V)=[c2/(- npb K)]-1/ (-1-n).

If b varies with a positively skewed distribution around the central value of 0.3, and all other parameters are as in scenario 6, then the results are as shown for scenario 8 in Table 3. Mean income, utility and relative variation of income do not change from scenario 6 to 8. There is some increase in the relative loss to disease in scenario 8 with either offtake rule.

Discussion

The costliest risk, in terms of foregone average income, is lack of feed caused by variable pasture production (Table 3). The variability of feed supply cannot be economically compensated by such measures as forage reserves and supplemental feed.

The loss caused by the variability of domestic livestock prices and high animal disease was about 54% of the absolute loss caused by variable pasture production.

Incorporating the variance and skewness of income into the analysis did not change the results dramatically. The utility measure based on wealth and the mean, variance and skewness of income, as compared to utility based only on mean income and wealth, was about 10% less with the target offtake rule and 4% with the feed equilibrium rule (Table 3).

Efficiency of offtake rules

The efficiency of the two offtake rules was very similar. The target offtake rule gave slightly higher income, less income variability, and higher utility than did the feed equilibrium rule. These offtake rules - if they were not already evident to experienced herders anyway - could not form the basis of new management practices to be recommended to producers by the extension service.

One interesting finding is that the conservative target offtake strategy did not make income substantially less variable than the feed equilibrium strategy. Another is that risk-aversion made less difference to the feed equilibrium strategy than to the target offtake strategy, apparently because the distributions of profits with the latter offtake rule were less negatively skewed.

Incentive problems

Incentive problems resulting from producers' biased subjective estimates of the likelihood of stock disease can lead to inadequate preventive treatment of those animals. Inadequate prevention of some stock could, in turn, cause more widespread and severe outbreaks of disease than would occur if all stock were treated correctly.

The results presented here indicating a sharp difference in economic results between low and high vector pressure situations suggest that there is potential for an incentive problem of this type to occur. In particular, if a sudden outbreak were to occur - for example, if vector pressure rose suddenly - then it is unlikely that all producers would react with the same celerity. Therefore, to prevent damage to livestock productivity, preventive treatment would have to be mandatory or sustained extension campaigns would be needed to raise producers' awareness of disease costs. Preventive treatment or extension would probably require substantial costs which would have to be subsidized because producers would fail to buy the economically optimal amount.

Future research

The paper has probed the probable consequences of variations in pasture production, output price, and livestock morbidity caused by disease on returns to livestock production and on the demand for veterinary inputs. Because the African data on livestock productivity are so sparse,* we lack good statistical estimates of production or cost functions for cattle and other stock, including the effects of such factors as disease, feed and management. A further problem is that producers lack complete knowledge of the effects of veterinary inputs not only on average output but also on its variance and skewness illustrate this problem in a study of California dairying.

* It was originally intended to use the data in Itty (1992) to estimate a production function for cattle with veterinary inputs and labour, among others, as explanatory variables. However, even that data, which are from the ATLN were not useful for that purpose.

Difficult incentive problems arise from the distributions of vector attack and disease transmission. In this paper, for lack of information about a more appropriate specification, it was assumed that those probabilities were not affected by the producers' actions. This assumption may be weak if either of those probabilities can be affected by the stocking rate. If the stock rate does affect the likelihood of attack or of subsequent infection, then not only must the decision of the individual herd owner about the intensity of production be taken into account, but that of all other herd managers likely to use the same common grazing areas must be considered too. Those decisions would greatly complicate economic analysis of the optimal level of disease control efforts to deploy because then the consequences of individual producers' decisions would not be independent.

With respect to disease, note that the low disease pressure scenarios could result from spraying to control vectors. This is likely to bias producers' subjective estimates of the severity of outbreaks, e.g. they would make producers initially misinterpret a high pressure situation as a low one. Second, the sudden outbreak of disease is not likely to self-correct via the mechanism of rising prices. This is because rising offtake leads to falling prices in the short run, making the return to veterinary inputs low. Third, the long term development of feed production, by reducing the cost of livestock production, can also reduce the unit price, thereby cutting the derived demand for veterinary inputs.

References

ANDERSON, J.R., DILLON, J.L. and HARDAKER, B. 1977. In: Agricultural Decision Analysis. Ames: Iowa State University.

BINSWANGER, H.P. and McINTIRE, J. 1987. Behavioral and material determinants of production relations in land-abundant tropical agriculture. Economic Development and Cultural Change 36 (1): 73-99.

DAHL, G. and HJORT, A. 1976. Having Herds: Pastoral Herd Growth and Household Economy. Stockholm: University of Stockholm Studies in Social Anthropology.

ITTY, P. 1992. Economics of village cattle production in tsetse-affected areas of Africa. Zurich: Unpublished Ph.D. dissertation. Swiss Federal Institute of Technology.

McINTIRE, J. 1991. Pastoralism and risk. In: Holden, D., Hazell, P. and Pritchard, A., eds. Risk in Agriculture. World Bank.

McINTIRE, J., BOURZAT, D. and PINGALI, P. 1992. Crop-Livestock Interaction in Sub-Saharan Africa. World Bank.

Von KAUFMANN, R., McINTIRE, J. and ITTY, P. 1990. Bioeconomic Herd Model for Microcomputer: User's Manual and Technical Reference Guide. Addis Ababa: ILCA.

Potential for modelling ecological responses to the control and prevention of disease in African livestock populations

M.B. Coughenour

Natural Resource Ecology Laboratory
Colorado State University
Fort Collins. Colorado. USA

Abstract
Introduction
When technological interventions overcome ecological constraints
Modelling ecological effects in relation to carrying capacity
Ecosystem modelling: needs and capabilities
Modelling interactions between livestock, humans and wildlife
Previous experiences using ecosystems analysis in African livestock and wildlife ecosystems
Conclusions
References


Abstract

The environmental consequences of livestock disease are increasingly being taken into account by policymakers and livestock disease specialists. Ecological modelling can help assess how livestock disease suppression may indirectly affect the environment through increases in livestock and human populations and associated grazing pressures. It can also help in assessing ecological constraints on livestock production, which affect the economic benefits of disease control. Ecological modelling can be used to assess environmental responses as well as ecological constraints.

Ecological modelling has advanced over the last decade due to technological as well as scientific progress. Spatially explicit models are now used at landscape to global spatial scales. Models are now readily integrated with geographical information systems and remote sensing.

Land-use models must be used to predict land conversions to livestock-based agriculture. Ecological models must then be used to model vegetation and soil responses to human and livestock utilization. Agro-ecosystem models can be used similarly. Finally, the predicted alterations in vegetation and soils must be used in models of wildlife habitat suitability, movements, and spatially and non-spatially structured population dynamics.

Introduction

Sustainable development is the development that meets the needs of the present without compromising the ability of future generations to meet their own needs - World Commission on the Environment and Development

With increasing emphasis being placed on ecologically sustainable economic and agricultural development and the rapidly increasing human populations in Africa, the potential ecological consequences of livestock disease control are likely to be taken into much greater account than they have in the past. Humans are only now realizing that producing food, providing shelter and increasing their comfort can have important side effects.

There are many potential environmental consequences of livestock disease control. Livestock disease strongly limits livestock and thus human populations throughout Africa. The hoped for result of livestock disease control is that livestock populations will increase. Unfortunately, this may have undesirable side effects like overgrazing (Sinclair and Fryxell, 1985; Bosch, 1989), soil deterioration (Lal, 1988; Graetz, 1989), competition with wildlife species and alteration of wildlife habitat (e.g. Talbot, 1972; Myers, 1973; Coe, 1980; Williamson et al., 1988). Many of Africa's national parks may owe their existence to the fact that tsetse makes them unsuitable for livestock (Coe, 1980). There are potential side effects of vector control programs that may arise from frequent herd movements to dips, bush clearing, burning and spraying. Along with increased livestock populations come elevated human populations. In Africa, this may imply increased levels of wood harvesting for fuel and construction. In productive areas, wildlife habitats will be converted to croplands. Finally, humans may elect to control or even eliminate wildlife populations in surrounding wildlands to eliminate disease vectors and prevent crop depredation. In the long run, negative ecological side effects have negative economic consequences as the ability of land to support grazing, cultivation, forestry and tourism declines.

In addition to improving our understanding of environmental consequences, ecological modelling may play useful roles in assessing the costs and benefits of disease control. Ecological modelling can be used to assess ecological constraints on livestock productivity. Calculations of the potential benefits of disease control depend upon correct assumptions of the maximum potential productivity that can be realized after the disease has been controlled, which in turn is affected by levels and dynamics of forage productivity. Also, livestock disease vector distributions and abundances are determined by vegetation as well as by climate.

The purpose of this paper is to explore the possibilities for using modelling to examine the likelihood of success of livestock disease control programs. If ecological costs and benefits can be incorporated into these considerations, implementation strategies of disease control programs could become more effective, more strategic, and more sustainable. Ecological modelling capabilities have advanced considerably over the last two decades. It is much more feasible now than it was a decade ago to use models to help assess the environmental consequences of livestock diseases and their control. I briefly discuss the needs for considering environmental consequences of disease control, then identify modelling procedures that can be used, drawing in part from experience gained modelling a pastoral ecosystem in Kenya.

When technological interventions overcome ecological constraints

In ecosystems that have not been affected by technological development, native grazers and browsers tend toward the natural 'carrying capacities' of their environments set by vegetation, water availability, predation, disease or other limiting resources. Periodic droughts or even long dry seasons act as 'bottlenecks' that keep populations well below levels that fully exploit vegetation resources. Predators and disease add further constraints on population rates of increase. Undeveloped pastoralism is limited by many of the same constraints.

Technological intervention into livestock-based human ecosystems tends to relieve these constraints, pushing populations to become limited only by forage and subsidies from outside the system. Water limitations are overcome by the development of wells, boreholes and reservoirs. Simple technological interventions by pastoralists like the protection of livestock from predators and shepherding to forage may cause livestock populations to be higher than their wildlife counterparts (however, wild ungulates in the Serengeti number over two million and seem to be food-limited rather than predator-controlled). Interactions with regional and national markets permit destocking during periods of low production, but they also permit more rapid and complete restocking. Energy subsidies to the system such as fertilizer and energy used to cultivate grain for livestock feed tend to boost population growth. It is easy to see that supplemental feeding of grain to a herd will relieve the natural forage limitation set by pasture or rangeland plant productivity.

Controlling livestock diseases where they currently have significant impacts on livestock mortality will likewise tend to force populations to levels where forage, and thus naturally regenerating resources, become limiting. Indeed, it is possible with technology to push livestock populations beyond their sustainable carrying capacities.

The risks of technological intervention are particularly acute on technological 'frontiers' like the American west at the turn of the century or the African Sahel from the 1950s onward (Sinclair and Fryxell, 1985; Le Houreou, 1989). At the frontier, natural resource levels are relatively high, thus promoting population expansion in the short-term at the expense of slow declines in natural resources over the long-term. In the Sahel there was widespread water development, which along with ample forage allowed the development of higher livestock densities. Eventually, however, the system was run down in many places. Indeed, much of Africa today is a technological frontier.

Ecological models that attempt to predict the potential consequences of livestock disease control must be able to describe the effects of these ecological constraints on current livestock populations. In much of Africa this implies that models must represent the effects of water availability and disease on herd density, as well as the effects of forage limitation. When diseases are controlled, it is likely also that other constraints will also be relieved through water development, improved marketing infrastructure, greater feed imports, and others. If these associated developments are ignored, the effects of disease control are likely to be underestimated.

Similarly, Geerling et al. (1986) suggested that integration of ecology in development requires 1) quantification of natural resources flows in terms of carrying capacity, 2) identification of natural ecological regulating factors in the natural and the developed system, 3) assess interventions to counter the natural regulating factors and potential side effects, and 4) define measures to integrate interventions with socioeconomic system to achieve balanced development.

Modelling ecological effects in relation to carrying capacity

Ascertaining ecological responses to livestock disease control can be compared to the problem of determining ecological carrying capacity for livestock and humans. If disease impacts on herd sizes can be predicted, then it is possible to at least assess whether there is any risk that carrying capacity will be exceeded.

In simple terms, livestock carrying capacity is the number of animals that can be supported over the long-term to achieve some dynamic equilibrium amongst soils, plants and animals. The amount of forage that can be sustainably produced under grazing or browsing is therefore central to carrying capacity calculations. While carrying capacity calculations are often based upon a fixed or average quantity of resources which then determines the number of resource users that the area can sustain (e.g. Coe et al., 1976; Kalff et al., 1985), carrying capacity is dynamic (e.g. Geerling and de Bie, 1986). In arid and semiarid regions, carrying capacity varies greatly within and among years due to plant responses to rainfall variation and time lags in plants responses. Dynamic simulation models can represent these variations, as well as subsequent population responses.

The concept of carrying capacity can take on varied meanings according to the objectives of land use (Geerling and de Bie, 1986; Coughenour and Singer, 1991). If the objective is to maximize wildlife diversity, the definition will be different than if the objective is to maximize energy and nutrient transfer to human populations. In production-oriented livestock systems, carrying capacity may be the number of animals that can be supported to attain maximal sustainable production over the long-term. Extinction of a minor plant species may be tolerated because its contribution to productivity is believed to be negligible, or is not understood. In pristine nature preserves, carrying capacity is defined by a naturally regulated animal population level. Plant species may persist because they evade herbivory or because herbivore populations do not increase to threatening levels. Thus, the potential for conflicts of interest between maximizing sustainable production and maximizing biodiversity become readily apparent.

Carrying capacity is actually a continuum along a gradient of what is demanded from an ecosystem. At the least, carrying capacity should be defined as the maximal stocking rate that ensures long-term agro-ecosystem sustainability. The most demanding definitions would arise where there is conservation of pristine wildlife ecosystems that harbour large reserves of biodiversity. In between, there is a wide range of acceptable livestock population levels. However, it is much more difficult to predict the consequences of a continuum of livestock densities than it is to simply predict a single valued carrying capacity that meets a single set of resource management objectives.

Thus, predictions of long-term agricultural and ecological sustainability in relationship to variations in livestock abundance are needed to assess the full range of ecological impacts of livestock disease control. These predictions require a synthesis of the long-term effects of livestock on forage plant production and survival, the effects of human wood use on woody plant populations, the direct and indirect effects of livestock on soil structure and fertility, and the secondary impacts on wildlife arising from competition for forage or alteration of habitat. The calculations must take into account temporal and spatial variations in carrying capacity due to climatic variability and landscape heterogeneity.

Ecosystem modelling: needs and capabilities

Livestock effects on plants and soils, and human effects on plants, soils and wildlife involve indirect as well as direct effects (Figure 1). For example, direct effects on plants are ramified into soil responses. Livestock and human impacts on vegetation have consequences for wildlife habitats. These interactive responses involve processes of plant growth, soil moisture dynamics and ecosystem nutrient cycles. Interactions and feedbacks among these processes occur at the ecosystem level of organization. Ecosystem studies take into consideration abiotic as well as biotic interactions, including the flows of water, carbon, inorganic nutrients and energy among soils, vegetation and animals. Indeed, an ecosystem is defined as an assemblage of abiotic and biotic components comprising an interdependent system (Tansley, 1935).

Models must be capable of simulating plant responses to various levels of grazing pressure. Plants generally respond positively or neutrally to light and moderate grazing and negatively to heavy grazing. Livestock and human impacts involve ecosystem level interactions among grazers, plants, soil and climate. There are several mechanisms involved in the response; alteration in photosynthetic leaf area, changes in photosynthesis rate, reductions in rates of soil water use by plants, increased rates of nutrient recycling by herbivores, meristems mortality arising from trampling and uprooting and other processes. Some responses are due to effects on plant numbers and sizes that arise from altered mortality and recruitment rates. Sustained overgrazing eventually results in soil deterioration. As plant productivity is diminished, carbon inputs to the soil decrease and soil organic matter declines along with soil fertility. Lack of vegetation cover may expose soils to direct raindrop impact, which can cause soil capping, increased runoff and finally erosion. These negative effects on soil fertility and water balance may induce a positive feedback cycle, as plant growth is further decreased, which then causes further soil degradation.

Figure 1. Direct and indirect interactions among livestock disease, livestock populations, ecological processes and humans. Note that livestock populations both affect and are affected by human land-use patterns.

Ecological modelling can also be used to disentangle ecological constraints on livestock production. Natality, mortality and resultant livestock herd dynamics may be simulated in response to animal condition, which depends partially upon forage intake rate. Disease effects can be modelled independently through their effects on animal weight gains, forage intake rates or mortalities. In most rangeland ecosystem models, the rate of forage intake by livestock depends upon the standing stock and nutrient content. Forage quantity and quality are predicted using plant growth models, taking into account water and nutrient limitations. Livestock growth is often simulated based upon the balance between energy or nitrogen forage inputs and metabolic losses. Energetic losses incurred by travelling to water and lactation are added to costs of basal metabolism (e.g. Coppock et al., 1986).

Mathematical models of ecological interactions predate the computer age by decades. Computer simulation models were first applied in ecology in the 1960s (e.g. Watt, 1968; Bledsoe and van Dyne, 1969; Odum, 1971). Computers revolutionized ecological modelling by enabling simulations of ecological interactions which were too complex to solve analytically. Modelling was increasingly employed by systems ecologists after that. Between 1970-1975, large ecosystem modelling programs were supported by the International Biological Program (IBP). Models were developed for sites in the grassland (e.g. Van Dyne, 1972; Innis, 1978) as well as other biomes. Ecosystem models continued to be developed after IBP.

The modeller's working environment is much more advanced now than it was ten years ago. Until about 1983, models were generally programmed using punch cards and run in batch mode on mainframe computers with relatively large memory capacities but slow processing speeds. The first microcomputers such as the 8086 and 80286 class machines available during the mid and late 1980s were still too slow and memory limited to be used for serious ecosystems modelling. Minicomputers were used extensively during that time while modellers moved from batch-oriented to interactive, terminal- or workstation-based environments. Now, however, serious ecosystem models can be developed and run on 80386 and 80486 class micros, as well as more sophisticated workstations. Software has also advanced dramatically over the last several years. Using 32-bit compilers, models having huge memory demands can be implemented. Skilled modellers can now develop or customize models in a small fraction of the time required a decade ago.

Increases in computer speed have made long-term and large-scale simulations feasible. Earlier ecosystems models were typically executed for one- to five-year time periods due to limitations in computer speed. Now, daily time step models can simulate decades of real time while weekly and monthly models routinely simulate centuries. Monthly and annual simulations of thousands of years are possible. Until recently, ecosystem simulations were limited to points or individual sites. Spatial heterogeneity plays a significant role in ecosystem function, and must be taken into account in ecological models, particularly for large spatial areas. Only several years ago did ecosystem modellers attempt to develop spatial simulation models. Spatial models are considerably more demanding of processor speed because model calculations must be performed at each and every spatial location. While a few large spatial models can only be run on supercomputers, most spatial simulation models can now be executed on micros and workstations as well.

The development of spatial data analysis systems has paralleled the above chain of events, also due to advances in computer technology. Many of the early hopes of remote sensing science during the late 1960s and 1970s are now being realized because it is now possible to store and process large amounts of data quickly and effectively. User friendly geographical information systems (GIS) are revolutionizing natural resource sciences and applications.

Recent developments, now ongoing, are the linkages of systems modelling with GIS and remote sensing, and the development of user friendly interfaces for models. There are but a handful of successful ecosystem model-GIS linkages at this time, but that is rapidly changing. Only now are there being developed graphical 'point and crick' user interfaces that command linked systems of ecosystem models and GISs.

There have been many important advances in the field of computer-based decision support systems (DSS) (Stuth and Lyons, 1993). DSS use heuristic approaches to solve complex problems involving integrations of ecological and economic factors. A wide range of tools and information sources, such as simulation models, expert systems, databases, GIS and remote sensing, are integrated. The design and construction of DSS are all oriented to the needs and ease of use of an end user (Stuth and Stafford-Smith, 1993).

Modelling interactions between livestock, humans and wildlife

Interactions between livestock, humans and wildlife can be modelled by representing vegetation responses to livestock and human utilization (Figure 2) and subsequent wildlife responses to the associated alterations of their habitats (Figure 3). The spatial distributions of wildlife habitats and extensive livestock systems impact the sustainabilites of both (Coughenour, 1991c).

However, how do we model patterns of land-use change? Human land-use impacts are particularly problematic for ecological modellers as it is really beyond the traditional bounds of their discipline. While geographers have developed models of human population distributions (e.g. Thomas and Hugget, 1980), there has been little interaction between ecologists and geographers. Two fundamentally different approaches to predicting land use might be recognized. The first is purely geographic and would predict changes in land use from environmental and economic variables and location theories. The second approach would be to predict livestock population dynamics at large spatial scales based upon environmental constraints, and then base land-use changes upon changes in livestock abundance as well as other geographic considerations. In reality, livestock abundance and land use are interactive, with causality flowing in both directions (Figure 1).

Maps of potential land use in Africa such as those developed by the FAO can possibly be used as a basis for predictions. However, it is critical to obtain data for actual land use. Ecological, disease and other constraints on land use are likely to cause substantial differences between actual and observed land use. It would, however, be highly instructive to compare potential and actual land use in relation to prior livestock disease controls and other factors. This would provide useful information for the development of land use change models.

Figure 2. A major role of ecosystem models is to calculate the effects of livestock and human resource utilization upon vegetation structure and function. Ecosystem models represent interactions among plant populations, biomass production, hydrology and nutrient cycling.

Possibly, a rule-based modelling approach could be taken to represent land use changes. Rule-based modelling, a branch of artificial intelligence, encodes knowledge into logical if-then relationships. Effects may be probabilistic related to causes. Probabilities of land-use change could be functions of environmental variables such as rainfall and soil fertility as well as economic and geographic factors. A wide range of outcomes may be possible, depending on which rules are stochastically invoked; however it may be desirable to track only the most probable outcome.

A further requirement is to predict the secondary ecological responses to either livestock increases or land-use change or both. Ecologists can readily predict potential vegetation responses to climate variables like rainfall, potential evapotranspiration, growing season length and temperature (e.g. Cramer and Leemans, 1993; Emmanuel et al., 1985). However, this approach does not take into account human land-use impacts nor does it model transitional vegetation types, dynamics or new types of vegetation (i.e. new combinations of plant types). Again, a rule-based approach could be employed to account for human impacts. Rules would have to be developed from existing data bases that quantify relationships between rates of forest clearing, burning, cultivation, road building, etc. and increases in livestock-based human populations and land usage. The rates of land use and vegetation change may also be predicted to enable scenarios to be created at various points in time into the future. Rates of deforestation can be calculated based on human wood demands and rates of wood production by trees. The latter may be predicted from rainfall and other variables, however such predictions are limited because wood production rates also decline as tree population densities decrease.

Figure 3. Predictions of vegetation responses from ecosystem models are used as inputs to models of wildlife habitats, distributions and populations. Wildlife includes plants as well as animals. Human land-use models provide inputs to wildlife habitat suitability as well as vegetation models. Humans may directly impact wildlife populations by hunting, culling etc.

Responses to human land-use practices can also be modelled more explicitly using process-oriented ecosystem models as discussed in the previous section. Plant growth rates and population dynamics and their responses to utilization (e.g. Coughenour, 1984, 1991a; Fulton, 1991) are simulated, along with their interactions with water budgets and nutrient (N, P) cycles (e.g. Parton et al., 1987).

Modelled vegetation responses are used as inputs to models of wildlife habitat suitability. Habitat suitability models have been developed and used extensively during the last decade (Verner et al., 1986). Habitat attributes such as vegetation type and abundance, topography and distance to water are used to assess the suitability of a habitat for various wildlife species. These models are usually based on observed correlations between habitat variables and the distributions and abundances of wildlife species.

An important habitat variable is the amount of forage biomass available to wild herbivores. Competition between livestock and wild herbivores may be explicitly modelled by simulating decreases in forage standing crops due to their respective forage offtakes. Foraging models then respond by predicting reduced rates of forage intake. In a more simplified approach, the surplus forage after livestock consumption could be calculated and then used as an input to a habitat suitability model.

The increasing emphasis that is being placed on conservation biology and biodiversity at regional, national and global spatial scales is beginning to yield new techniques for assessing wildlife distributions and critical wildlife areas. Geographic techniques are used to map biodiversity from species inventory data and range maps. Analyses are then performed to optimally locate wildlife preserves to maximize biodiversity. Within the US Fish and Wildlife Service the GAP analysis program has emerged (Scott et al., 1993). This is a geographic inventory and analysis of critical gaps in the distribution and abundance of wildlife species and their habitats. Other similar techniques were reviewed by Pressey et al. (1993). A limitation of these approaches is that they are static rather than dynamic. They characterize potential habitat value, but do not examine population dynamics or animal movements to arrive at the potential.

As human populations increase, it becomes increasingly difficult to set aside large, intact natural ecosystems for conservation. Instead, human land use must be managed to minimize negative impacts. In order to accomplish that, it is necessary to determine what population sizes or densities are needed to ensure long-term persistence or viability. Conservation biologists estimate minimum viable population sizes or densities (MVPs) in cases where the biology of a species is well understood. A number of approaches have been used to estimate MVPs (Soule, 1987; Boyce, 1992). Generally, simple population dynamics models are employed. Models are run repeatedly, with stochastic environmental variations in birth and death rates among runs that reflect the range of variation likely to be encountered due to climate variation, disease outbreaks and other catastrophes (as in Monte Carlo modelling [see Morris, this volume]). In some cases the effects of spatially structured metapopulations, with stabilizing dispersal and movements among subpopulations, are considered. These are risk analyses in that they attempt to determine risks of extinctions in probabilistic terms (e.g. a 95% probability of persistence over 100 years) (Burgman et al., 1993).

These techniques can be usefully merged with assessments of current and projected livestock distributions and associated land conversions to cultivation and pasture. The linkage between those projections and potential impacts on biodiversity will be problematic, however. Qualitatively, it is often assumed that livestock development will have negative effects on wildlife, but quantitative assessments are lacking. Furthermore, impacts of livestock development fall along a continuum - impacts can be minimized through proper land-use planning, herd management and resource management tactics and their environmental consequences must be included in models of the environmental consequences of livestock disease and disease control.

Previous experiences using ecosystems analysis in African livestock and wildlife ecosystems

Applications of an ecosystem approach have been rare in studies of African livestock or wildlife. Although ecosystem level analyses have been conducted for over two decades, many of these exercises have been academic or have not considered humans. Exceptions are modelling studies of pollution impacts (many), agricultural practices (e.g. Cole et al. 1989), the works of H.T. Odum (1971), global modelling studies (Meadows et al., 1972) and human ecology studies (Little et al., 1984, Moran, 1987; Weinstein et al., 1983).

The South Turkana Ecosystem Project (STEP) in northwest Kenya employed an integrative, multidisciplinary ecosystems approach to understand the structure and functioning of a nomadic pastoral ecosystem. STEP, funded by the US National Science Foundation, conducted research on pastoral ecology in the southern portion of Turkana District, Kenya, over a 13-year period, beginning in 1980 (Little et al., 1984; Ellis and Swift, 1988; Coughenour et al., 1985, McCabe and Ellis, 1987). STEP linked together studies of climate, soils, hydrology, vegetation, livestock and human biology and behavior. Systems modelling was used throughout the project.

STEP studies focused on a 10,000 km² area of the southern portion of Turkana District, Kenya, that is topographically, geomorphically and climatically diverse. South Turkana is mainly arid (300-550 mm/yr rainfall), but also includes very arid (15-300 mm/yr) and semiarid (450-900 mm/yr) climatic zones. Vegetation is a diverse mixture of dwarf shrub grassland, dry thorn bushland, grassland and savanna. Pastoralists are nomadic, retaining a traditional subsistence lifestyle that depends on products of goats, sheep, cattle, camels and donkeys. Milk is by far the most important livestock product (Galvin, 1988). Pastoralists move several times per year in response to changing distributions of forage, water, security and another factors (McCabe, 1983; McCabe and Ellis, 1987).

It became apparent early in the research (ca. 1983) that a spatial data base was needed to organize information and conduct ecosystem analyses. Soils, landform, topography, hydrology, vegetation and climate maps were computerized using a GIS. Maps of seasonal and permanent wells and large rivers were used to develop maps of distance to water. A temporal sequence of monthly rainfall maps was generated for a seven-year period (1982-1988) from rainfall data gathered at irregularly spaced weather stations in the study area. Rainfall maps were then generated using spatial interpolation, accounting for effects of elevation.

Satellite remote sensing data were useful for verifying model predictions of green vegetation dynamics. Normalized difference vegetation index (NDVI) maps were derived from reflectance data from the AVHRR sensors on NOAA polar orbiting satellites (Tucker et al., 1985). The first data that NASA (C.J. Tucker) provided to us in 1984 were monthly NDVI images of 7.5 x 7.5 km pixels. Later 4.2 x 4.2 km data were provided. Time series of monthly NDVI images from 1980-1988 were used to verify ecosystem model predictions of the distribution and dynamics of green plant biomass.

Diverse studies of vegetation, livestock and human ecology were synthesized in a static energy flow model of the ecosystem (Coughenour et al., 1985). Spatial vegetation data were integrated into total ecosystem estimates of plant productivity. We were able to quantify the amount of energy fixed by plants, consumed by animals, and then transferred to humans. This modelling exercise forced us to identify interdependencies in the system in quantitative terms. We identified where the inefficiencies of energy transfer were located, how these transfers were ecologically constrained, and why these inefficiencies were ecologically significant. The analyses revealed significant differences between the structure and functioning of production-oriented systems such as the raising of livestock for meat and maintenance-oriented systems such as subsistence pastoralism.

A spatial, dynamic ecosystem simulation model (SAVANNA) was later constructed (Coughenour, 1991b, 1992). This model simulates soil water balance, plant growth, livestock foraging and livestock condition for 4.2 x 4.2 km grid cells. Simulated pastoral livestock are dynamically distributed over the landscape in relation to habitat suitability, which is modelled from the combined distributions of forage, water, topography and woody canopy cover. Data stored in the GIS database are used for model initialization. GIS is used to analyse and plot model output. Monthly rainfall maps are used as model inputs. We used the model to demonstrate the significance of spatially heterogeneous landscapes for ecosystem level function and we assessed the importance of livestock movements for ecological stability. Simulations demonstrated the significance of livestock movements (e.g. Coughenour, 1991c) and the effects of restrictions on livestock movement arising from livestock raiding by a neighbouring tribe.

A more developed version of the SAVANNA landscape model simulates plant population dynamics as well as biomass production for multiple life forms or functional groups of plants. The effects of livestock grazing and browsing on plants are mechanistically, but parsimoniously, represented (Coughenour, 1991a). Multiple livestock species can be simulated. The energy balance of livestock is explicitly modelled as are weight dynamics. Livestock population dynamics of five sex/age classes are simulated. Mortality and nasality rates are functions of livestock condition index, i.e. current body weight relative to minimum and maximum body weights.

SAVANNA is optionally implemented in a newly developed GIS-based graphical user environment (Oyasin Circle Solutions, Longmont, Colorado). Linkages between ARC/INFO data and the model are transparent. Point and click menu environments or 'shells' are customized to automatically perform complex procedures demanded by particular users.

We have coupled the ecosystem model to a model of pastoralist decision making based on an earlier model developed by D. Swift et al. (Ellis et al., 1990). A similar model was developed by Weinstein et al. (1983). Subsistence pastoralism is represented in terms of energy supply and demand from livestock and from other sources of food energy. A rule-based approach to pastoral decision making is taken. In times of livestock food shortage, a set of prioritized actions are implemented for making up the shortfall, ranging from slaughtering to hunting to famine relief. The model includes options for interactions with a market economy in which livestock are bought and sold. Cash derived from livestock sales may be used for buying livestock, food or non-food items according to prevailing circumstances. Currently, the model is being used to examine effects of different plans and scenarios of water resource development, marketing improvements and livestock mortality rates (G. Njiru, Ph.D., in progress).

The ecosystem modelling approach used in STEP would be directly applicable to assess effects of the control of livestock diseases in the semi-arid climatic zone where spatially extensive and relatively low input livestock production systems are common (e.g. large individual or group ranches, communal grazing area and transhumance). Human populations in these semi-arid climatic zones are expected to increase substantially.

There are some important differences between the subsistence pastoralism of arid Turkana and more sedentary production-oriented systems characteristic of semiarid and subhumid climatic zones. However, ecological models are generally applicable to a wide range of ecoclimatic zones. For instance, vegetation production and plant responses to grazing in SAVANNA are modelled in exactly the same fashion for arid and subhumid systems, using different parameterizations. The SAVANNA plant growth submodel has been used to simulate subhumid Serengeti tallgrasses (Coughenour, 1991a), as well as tree, shrub and grass species from the cold temperate zone in Canada.

While livestock in high production zones do not move as extensively as they do in Turkana, the SAVANNA livestock distribution modelling approach might be adapted to simulate reallocation of animals among ranches or farms or even regional scale distributions of sedentary livestock enterprises in relation to changes in livestock habitat suitability or localized excesses or deficiencies in herd sizes. Livestock would not be freely redistributed on a monthly basis as they are now in the Turkana version of SAVANNA; however, limited redistributions may simulate market exchange (destocking, restocking) responses to landscape, regional and national climate patterns as well as gradual changes in constraints on land use such as those arising livestock disease control. More sophisticated marketing and geographic models may come into play later.

In intensive livestock production systems there may be high levels of investment in farm infrastructure and maintenance, and significant cash outlays may occur for the purchase of animal feed. Thus, a farm or ranch level livestock management model with appropriate accounting procedures and management options (e.g. Stafford-Smith and Foran, 1990; Morris, this volume; Wight and Skiles, 1987; Mukhebi et al., 1989) should be used in place of the pastoralism submodel that was used in the Turkana modelling work.

Conclusions

We now have the capacity, using models, to make reasonably informed predictions about long-term ecological responses to African livestock development. We can use models to estimate changes in soil fertility, tree cover and herbaceous and woody biomass production in response to livestock production systems and climate over 10 - 100-year periods, accounting for the effects of spatial variations in climate, soils, topography and vegetation over landscapes, regions and the continent.

It would be useful to begin applying models to a few landscapes or subregions where there are sufficient data to parameterize and test the models. These data would include climate, soils, vegetation, topography, wildlife distributions and human population density and land use. Optimally, the study area(s) would encompass a sufficiently wide range of climate, soils, vegetation and land uses to develop an understanding of how these factors affect ecological responses. Longitudinal data on ecological and human responses over time are equally important. Thus, it will probably be necessary to capitalize upon prior research programs.

Maps of maximum sustainable livestock densities for landscape or regions may be developed or the models may be used to analyse the spectrum of responses and their sensitivities to specific variables. The use of ecosystem models to assess livestock production and disease responses to ecological constraints should not be overlooked. This information is useful in estimating potential sustainable livestock production at various levels of investment in disease control and where disease control will be most effective.

Spatially implemented ecosystem models like SAVANNA or CENTURY (Parson et al., 1987) can therefore address many of the potential environmental consequences of livestock disease control. They can address the side effects of livestock population increases by simulating plant and soil responses to herbivory, burning, brush clearing, wood cutting and land-use change. Potential overuse of areas near dips can be represented by modelling livestock density distributions. Simulated soil and vegetation responses can then be used in models of wildlife habitat suitability. The responses of wildlife populations to these changes could be estimated by using the outputs of the ecosystem models as inputs to wildlife habitat suitability models. In some cases population models would be employed to estimate wildlife population dynamics and minimum viable population sizes or densities. Modelling in natural resource management, as in other fields, should be viewed as part of an ongoing process. In adaptive resource management (Holling, 1978; Walters, 1986), managers continually learn about ecological responses to their actions by employing a wide range of tactics and carefully observing the results. Simultaneously, models become increasingly robust as they are tested, fail and are then improved. Over the long-term (decades), iterations between modelling, research and management converge on optimal solutions. While models can and should be used as tools for decision support, many of the most important benefits of models are derived from participating in the modelling process.

References

BLEDSOE, L.J. and van DYNE, G.M. 1969. Evaluation of a digital computer method for analysis of compartmental models of ecological systems. Technical Report ORNL-TM-2414. Oak Ridge, Tennessee: Oak Ridge National Laboratory.

BOSCH, O.J.H. 1989. Degradation of the semi-arid grasslands of southern Africa. Journal of Arid Environments 16: 165-175.

BOYCE, M.S. 1992. Population viability analysis. Annual Review of Ecology Systematics 23: 481-506.

BURGMAN, M.A., FERSON, S. and AKCAKAYA, H.R. 1993. Risk Assessment in Conservation Biology. London: Chapman and Hall, 314 pp.

COE, M.J. 1980. African wildlife resources. In: Soule, M.J. and Wilcox, B.A., eds. Conservation Biology: An Evolutionary Perspective. Sunderland: Sinauer Association, pp. 273-302

COE, M.J., CUMMINGS, D.H. and PHILLIPSON, J. 1976. Biomass production of large African herbivores in relation to rainfall and primary production. Oecologia 22: 341-354.

COLE, C.V., STEWART, J.W.B., OJIMA, D.S., PARTON, W.J. and SCHIMEL, D.S. 1989. Modelling land use effects on soil organic matter dynamics in the North American Plains. In: Clarholm, M. and Bergstrom, L., eds. Ecology of Arable Land. Kluwer Academic Publishers, pp. 89-98.

COPPOCK, D.L., SWIFT, D.M., ELLIS, J.E. and GALVIN, K. 1986. Seasonal patterns of energy allocation to basal metabolism, activity and production for livestock in a nomadic pastoral ecosystem. Journal of Agricultural Science 107: 357-365.

COUGHENOUR, M.B. 1984. A mechanistic simulation analysis of water use, leaf angles, and grazing in East African graminoids. Ecological Modelling 26: 203-220.

COUGHENOUR, M.B. 1991a. Dwarf shrub and graminoid responses to clipping, nitrogen and water: simplified simulations of biomass and nitrogen dynamics. Ecological Modelling 54: 81-110.

COUGHENOUR, M.B. 1991b. A GIS/RS based modelling approach for a pastoral ecosystem in Kenya. In: Resource Technology 90-Second International Symposium On Advanced Technology in Natural Resources Management, Proceedings. Bethesda: American Society of Photogrammetry and Remote Sensing.

COUGHENOUR, M.B. 1991c. Spatial components of plant-herbivore interactions in pastoral, ranching and native ungulate ecosystems. Journal of Range Management 44: 530-542.

COUGHENOUR, M.B. 1992. Spatial modelling and landscape characterization of an African pastoral ecosystem: a prototype model and its potential use for monitoring drought. In: McKenzie, D.H., Hyatt, D.E. and McDonald, V.J., eds. Ecological Indicators, Vol. I. London and New York: Elsevier Applied Science, pp. 787-810

COUGHENOUR, M.B. and SINGER, F.J. 1991. The concept of overgrazing and its application to Yellowstone's northern range. In: Keiter, R. and Boyce, M., eds. The Greater Yellowstone Ecosystem: Redefining America's Wilderness Heritage. New Haven: Yale University Press, pp. 209-230.

COUGHENOUR, M.B., ELLIS, J.E., SWIFT, D.M., COPPOCK, D.L., GALVIN, K., J.T., McCABE and HART, J.T. 1985. Energy extraction and use in a nomadic pastoral ecosystem. Science 230: 619-625.

CRAMER, W.P. and LEEMANS, R. 1993. Assessing impacts of climate change on vegetation using climate classification systems. In: Solomon, A.M. and Shugart, H.H., eds. Vegetation Dynamics and Global Change. London: Chapman and Hall, pp. 190-217.

ELLIS, J.E. and SWIFT, D.M. 1988. Stability of African pastoral ecosystems: alternate paradigms and implications for development. Journal of Range Management 41: 450-459.

ELLIS, J.E., GALVIN, K.A., McCABE, J.T. and SWIFT, D.M. 1990. Pastoralism and Drought in Turkana District, Kenya. A Report to NORAD. Bellvue: Developmental Systems Consultants, Inc., 202 pp.

EMMANUEL, W.R., SHUGART, H.H., STEVENSEN, M.P. 1985. Climatic change and the broad scale distribution of terrestrial ecosystem complexes. Climatic Change 7: 29-43.

FULTON, M.F. 1991. A computationally efficient forest succession model: design and initial tests. Forest Ecology and Management 42: 23-34.

GALVIN, K.A. 1988. Nutritional status as an indicator of impending food stress. Disasters 12: 147-156.

GEERLING, C. and De BIE, S. 1986. The concept of carrying capacity and land-use. Netherlands Journal of Agriculture Science 34: 339-347.

GEERLING, C., BREMAN, H. and BERCZY, E.T. 1986. Ecology and development: an attempt to synthesize. Environmental Conservation 13: 211-214.

GRAETZ, R.D. 1989. Desertification: a tale of two feedbacks. In: Mooney, H.J. et al., eds. Ecosystem Experiments. Chichester: Wiley and Sons, pp. 59-87.

HOLLING, C.S. 1978. Adaptive Environmental Assessment and Management. New York: Wiley and Sons.

INNIS, G.S., ed. 1978. Grassland Simulation Model. Ecological Studies 26. New York: Springer-Verlag, 298 pp.

KALFF, J., DOWNING, J.A. and SMITH, T.T. 1985. Rainfall, agriculture, livestock and human density in the dry regions of Kenya. Journal of Arid Environments 9: 173-183.

LAL, R. 1988. Soil degradation and the future of agriculture in sub-Saharan Africa. Journal of Soil and Water Conservation 43: 414-151.

Le HOUEROU, H.N. 1989. The grazing land ecosystems of the African Sahel. In: Ecological Studies, 75. Berlin: Springer-Verlag, 282 pp.

LITTLE, M.A., DYSON-HUDSON, N., DYSON-HUDSON, R., ELLIS, J.E. and SWIFT, D.M. 1984. Human biology and the development of an ecosystem approach. In: Moran, E., ed. The Ecosystem Concept in Anthropology. Boulder: Westview Press, pp. 103-131.

LUSIGI, W.J. 1981. New approaches to wildlife conservation in Kenya. Ambio 10: 87-92.

McCABE, J.T. 1983. Land use among the pastoral Turkana. Rural Africana 15-16: 119-126.

McCABE, J.T. and ELLIS, J.E. 1987. Beating the odds in arid Africa. Natural History 96 (1): 33-41.

MEADOWS, D.H., MEADOWS, D.L, PANDERS, J. and BEHRENS III, W.W. 1972. The Limits to Growth. New York: Universe Books, Signet Books, 207 pp.

MORAN, E., ed. 1987. The Ecosystem Concept in Anthropology. Boulder: Wesview Press.

MUKHEBI, A.W., WAITHANGA, S.P.J., PERRY, B.D., IRVIN, A.D. and MORZARIA, S.P. 1989. Financial analysis of East Coast fever control strategies on beef production under farm conditions. Veterinary Record 125: 456-459.

MYERS, N. 1973. Tsavo National Park and its elephants: an interim appraisal. Biological Conservation 5: 123-132.

ODUM, H.T. 1971. Environment, Power and Society. New York: John Wiley and Sons, 331 pp.

PARTON, W.J., SCHIMEL, D.S., COLE, C.V. and OJIMA, D.S. 1987. Analysis of factors controlling soil organic matter levels of grasslands in the Great Plains. Soil Science Society of American Journal 51: 1173-1179.

PRESSEY, R.L., HUMPHRIES, C.J., MARGULES, C.R., VANE-WRIGHT, R.I. and WILLIAMS, P.H. 1993. Beyond opportunism: key principles for systematic reserve selection. Trends in Ecology and Evolution 8 (4): 124-128.

SCOTT, J.M., DAVIS, F., CSUTI, B., NOSS, R., BUTTERFIELDS, B., GROVES, C. ANDERSON, H. CAICCO, S., D'ERCHIA, F., EDWARDS Jr., T.C., ULLIMAN, J. and WRIGHT, R.G. 1993. Gap analysis: a geographic approach to protection of biological diversity. Wildlife Monographs 123: 41 pp.

SINCLAIR, A.R.E. and FRYXELL, J.M. 1985. The Sahel of Africa: ecology of a disaster. Canadian Journal of Zoology 63: 987-994.

SOULE, M.E., ed. 1987. Viable Populations for Conservation. Cambridge: Cambridge University Press.

STAFFORD-SMITH, M. and FORAN, B. 1990. RANGEPACK: the philosophy underlying the development of a microcomputer-based decision support system for pastoral land management. Journal of Biogeography 17: 541-546.

STUTH, J.W. and LYONS, B.G., eds. 1993. Decision Support Systems for the Management of Grazing Lands: Emerging Issues. UNESCO, Paris and Parthenon Publishing Group Ltd., 301 pp.

STUTH, J.W. and STAFFORD-SMITH, M. 1993. Decision support for grazing lands: an overview. In: Stuth, J.W. and Lyons, B.G., eds. Decision Support Systems for the Management of Grazing Lands: Emerging Issues. UNESCO, Paris and Parthenon Publishing Group Ltd., pp. 1-36.

TALBOT, L.M. 1972. Ecological consequences of rangeland development in Massailand, East Africa. In: Farvar, M.T. and Milton, J.P., eds. The Careless Technology. Garden City, New York: Natural History Press, pp. 694-711.

TANSLEY, A.G. 1935. The use and abuse of vegetational concepts and terms. Ecology 16: 284-307.

THOMAS, R.W. and HUGGETT, R.G. 1980. Modelling in Geography: A Mathematical Approach. Totowa, New Jersey: Barnes and Noble Books, 338 pp.

TUCKER, C.J., TOWNSEND, J.R. and GOFF, T.E. 1985. African land cover classification using satellite data. Science 227: 369-375.

Van DYNE, G.M. 1972. Organization and management of an integrated ecological research program with special emphasis on systems analysis, universities and scientific cooperation. In: Jeffers, J.N.R., ed. Mathematical Models in Ecology. Oxford: Blackwell Publishers.

VERNER, J., MORRISON, M.L. and RALPH, C.J., eds. 1986. Wildlife 2000: Modelling Habitat Relationships of Terrestrial Vertebrates. Madison: University of Wisconsin Press.

WALTERS, C.J. 1986. Adaptive Management of Renewable Resources. New York: MacMillan.

WATT, K.E. 1968. Ecology and Resource Management. New York: McGraw-Hill, 450 pp.

WEINSTEIN, D.H., SHUGART, H.H. and GRANDT, C.C. 1983. Energy flow and the persistence of a human population: a simulation analysis. Human Ecology 11: 201-225.

WIGHT, J.R. and SKILES, J.W., eds. 1987. SPUR-Simulation of Production and Utilization of Rangelands. Documentation and User Guide. US Department of Agriculture. Agriculture Research Service. ARS-63.

WILLIAMSON, D., WILLIAMSON, J. and NGWAMOTSKO, K.T. 1988. Wildebeest migration in the Kalahari. African Journal of Ecology 26: 269-280.

Session discussion

The poor quality of existing data on the effect of diseases on production, the economic effect of diseases under different circumstances and the lack of data in several key areas enabling us to understand the complex interactions of the mechanisms involved was recognized to be a problem, even in developed countries. The example of trypanosomiasis was raised, in which not only the pathogenic effect of acute or chronic anaemia plays a role, but also the high protein and energy cost of replacing lost haemoglobin.

However, it was emphasized that methodologies both for field assessment of economic impact and for the broader scale assessments were available, and have been applied for some diseases. Examples presented were studies to assess the relative importance of diseases of livestock in Thailand, and the economic importance of ECF in eastern and southern Africa.

An example of the ranking of direct losses caused by diseases to the sheep industry in Australia was presented, and a broad estimate that diseases were accounting for 20-30% losses in current production was proposed. As far as trypanosomiasis was concerned, the importance of indirect losses was emphasized.

The discussion then focused on the type of economic analysis required. Although there is a political and donor need for absolute figures on the possible losses incurred by livestock industries, more realistic estimates are those that measure the extent to which disease impact can be reduced by control measures; assessment should be done on the basis of benefits to be achieved as a result of particular control options, not total losses incurred as a result of the presence of the disease.

The possible source of funding for these activities was discussed. As an example the role of the pharmaceutical industry in the development of the helminth models was highlighted. However, concern was expressed about the possible distortion that may result in the development of integrated control programs where all interests might not be served. The public sector may need to play a significant role, particularly in the early stages.


Previous Page Top of Page Next Page