Models and the analysis of productivity in extensive livestock systems in Israel

Introduction

There are no wide open ranges in Israel and not very many extensively managed herds of cattle or flocks of sheep. Possibly that is why we have so many models of pastoral systems. It seems that when development problems are particularly complex and essentially insoluble through benevolent intervention, the urge to challenge the obduracy of the system with a model becomes overpowering. In a sense, system modelling is a relatively easy way to avoid facing real problems. It seems as if almost anybody who really wants to build a simulation model and become a creator, albeit of a flimsy world of computer output.

In Israel models, particularly those of livestock systems, seem to have the property of spontaneous generation since the historic workshop on simulation modelling led by the late Professor George M. van Dyne in the early seventies. Even though no problems seem to be solved, few are deterred from going ahead and building a new model. Sometimes older models are cannibalised, often ignored (probably just as well) and the result is a glorious proliferation, especially for a community which produces only about 10% of its red meat consumption from herds based on the range.

Table 1 lists pastoral models in Israel, complex and simple, biological, economic and bio-economic, that have been published. Efforts were not limited to extensive herds and a promising analysis of dairy herd nutrition was discontinued because of in-house conflicts (Goldman et al, 1977; Talpaz et al, 1980). The list makes up a mixed bag which is rather difficult to review. In order to evaluate these models we need to define criteria for classification and possibly even evaluation.

Table 1. Extensive livestock production models in Israel.

Classification and evaluation

The time-tested criterion for biological taxonomy is sex, and it would be gratifying to be able to classify simulation models on such a universally useful basis. Unfortunately, this is not possible. A number of criteria can be used to classify models including the generality or specificity of the model, how widely applicable it is, and its complexity. Probably a more interesting approach would be to classify models by their objectives. That would also provide a recognisable measure of success, assuming that the objective was worthwhile in the first place.

Livestock systems models can serve a number of purposes, the most common of which are research, management and planning.

Research objectives of modelling aim at testing how well we understand the functional nature of complex systems in terms of known or hypothesised processes. A special aspect is the study of the sensitivity and stability characteristics of the system in question. The general assumption is that the functions that define the model are a good representation of the relevant characteristics of the system, so that analysis of the model has relevance to the physical system that is the subject of the study. We will see that when the objectives are clear and the model definition is short and concise, its use can lead to insights that can be classified as 'advance in knowledge'. The Israeli models that can be classified as research models, by nature or by declaration, are the NEGEV, KAHN, SIMPLE and INTAKE models.

The NEGEV model was a class exercise that started off in the van Dyne simulation workshop during 1971 and continued for more than a year at the Hebrew University and ARO. It was meant to show that, with a little training, simulation modelling of complex systems is feasible and can be done by a well-run interdisciplinary team and that it can produce workable tools that can be used for presumably useful system analysis. The NEGEV model had all the classical components of the IBP ecosystem models: Abiotic, primary producers, secondary producers, decomposers and managers. It described a sheep grazing system in the northern Negev based largely on data from Migda. The model indicated that when unmanaged and confined to a given area, sheep would die out in severe drought years, and that this could be avoided only by managerial adjustment of sheep numbers (Tadmor et al, 1977). No doubt, further work on the model could have produced a richer harvest of insights but it seems that the 'post-natal depression' that follows many massive modelling efforts had a deadening effect on further involvement. The decease of one of the moving spirits of the exercise certainly reduced motivation and NEGEV did not extend far beyond the bounds of the exercise it was originally intended to be.

The KAHN model is a fairly complex cattle simulation model that concentrates on describing animal performance, including both growth and reproduction, as a function mainly of feed quality, and of animal and breed characteristics. It is a development of the TAMU model, but is based on single animals rather than on classes in a herd and has been used mainly to validate some of the assumptions underlying the process definitions. It has been carefully constructed and has raised questions regarding the appropriateness of some of the equations used in the TAMU model. It can be used as a herd management tool, but to date has been used mainly for research. This model will be discussed in greater detail during this workshop.

The Noy-Meir SIMPLE model is a reductionist model taken to the extreme. Basically it defines only a growth function and a grazing (or consumption) function, in such a way that graphical and analytical methods of investigation can be used to solve problems of system stability. Certain problems like rotational grazing or seasonality, which cannot be treated analytically, are fairly easily solved by numerical methods.

By defining the problem as concisely as possible, these models have achieved a generality that has both added significantly to the understanding of grazing dynamics and provided a strong link between pasture management and classical population dynamics in ecology. The central issues of grazing system stability are treated in detail and provide a theoretical basis for rational pasture management. The SIMPLE model is our best example of an analysis that has attained a worthwhile scientific objective. It almost suggests that the impact of a model will bear an inverse relationship to its complexity. I say almost, because the analysis of the SIMPLE model is, conceptually at least, rather more sophisticated than that of many more complex models.

The INTAKE model is a mechanistic model which will also be discussed during this workshop. It is also a research model but even though it is concisely formulated, it is considerably more complex than the SIMPLE model. INTAKE mainly deals with behavioural aspects of intake and has been used to generate consumption functions. Problems of parameterisation and validation involve considerable experimental effort that cannot always meet model needs, especially if precision is important

The next set is the management models. These as a rule have practically-oriented management objectives. The management problems stem from the complexity of the interactions between climate, pasture, animal, nutrition and economic factors. Consequently, most of the management models are open-ended, in the sense that there is no conceptual limit to their complexity. Only the competence, patience and perseverance of the modeller set a limit. This group of models includes MIGS1, ZABAN, BENJAMIN, and BEEFX. UNGAR is also a management model, but is based on a more careful formulation and analysis of objectives.

The first attempt at setting up a full-blown agricultural system model resulted in the Migda system model, (MIGS1). It involved a concerted effort by a group of livestock and pasture specialists and, in fact, translated the current views on the functioning of the main components of the system into an operative continuous simulation model. By the time the model was completed, most of the problems that the model was meant to address were almost forgotten. Some new problems, such as timing of weaning, were analysed with the model and influenced the herd-management decisions taken at Migda, but only to a very limited extent. ZABAN and BENJAMIN are sheep husbandry models of the same genre. They too have been developed and analysed and have since remained on the shelf. Their effect on management practices has been no greater than that of MIGS1.

The UNGAR model, or rather set of models, will be discussed in more detail during the workshop. It treats separate problems that face the manager of agropastoral systems of the type developed at Migda. The problems are clearly and concisely defined and their nature exhaustively analysed, both analytically and graphically. Among the insights obtained is the division of the management space into areas of relative stability that are separated by areas, often rather narrow, where decisions are very sensitive to parameter changes or to data accuracy. Recognition of these different areas of robustness and instability can help to rationalise decision making under the conditions of uncertainty that are typical of the systems studied. Unfortunately, there has been little opportunity to implement the insights gained from this study, mainly because it has only recently been completed. Whether the impact of this model on management practices will be any different from its fore-runners remains to be seen. But if its application remains limited, it would, in no small measure also be due to the fact that the agropastoral systems which it is meant to serve are mainly experimental as yet and are only now being developed on a farm scale.

BEEFX can barely be called a model. It is essentially a feed and performance calculator that can produce an annual balance of a beef herd grazing a seasonal Mediterranean pasture. The feed requirements, pasture and feed values and performance parameters are all empirical table data, taken from the NRC manuals and from local farm data. Its special characteristic is that it is totally farm oriented and uses parameters that describe the herd, pasture, available feed, prices and performance standards as registered by the manager of the herd being analysed. When these data are available various management options can be tested for profitability or for production efficiency. The management options include hens size, cow size, feeding strategies, timing of breeding and weaning, supplementary feeding options and target sale weights as well as other sensitivity analyses. It calculates the evaluation parameters for one-year cycles and does not explicitly consider carry-over effects from one year to the next. Conceptually, the model is rather crude and does little more than what a rancher, concerned about improving his herd management, would do on the back of a match box. However, it does it on a much grander scale and allows much greater resolution than is normally possible with a heterogeneous herd and many management options. This model has been developed in close cooperation with the extension service and consequently has been introduced to a number of ranches. It has only recently been launched after an (over– ) extended period of development, but has provoked interest among extension workers and herd managers. It has already been used as an aid to determining whether to increase a herd and to what extent. But how much impact it will have on wider problems of herd management in Israel remains to be seen.

The group of planning models includes PSG and GRAPHICA. They are both input/output models that define a wide range of system conformations. They are also based on agropastoral systems appropriate to the northern Negev of Israel and were conceived as a means for determining appropriate planning and research objectives in fostering livestock and crop integration under dryland conditions. Both models can be divided into two main components: system definition and system evaluation. The system definition is a flexible accounting algorithm that defines the input/output relations of the system as a consequence of changing herd performance levels, pasture utilisation methods and other management parameters. The system evaluation component is basically different in both models. In PSG, the input/output model is combined with a multiperiod linear programming model to determine the optimum course of system selection over a development period, normally 15 years. The program takes into account regional resources and constraints and defines regional consumption as the basic target function. GRAPHICAL is a static model but can, however, be run over a series of years. Its main characteristic is that it evaluates the systems in a multiple-goal setting. The goals that have been defined are basically farmers' income, regional trade balance and regional employment.

The PSG linear-programming model has been analysed in some detail and has indicated the relative advantages and disadvantages of the three sheep breeds that are potentially useful in the area. GRAPHICAL has been applied to a number of management and planning problems and has helped clarify the system characteristics that would probably prevent widespread acceptance. It also indicates what the conditions are for acceptance and the direction for further research. Recently, PSG has also been developed into a multiple-goal model in a study planned to investigate feasible development scenarios in a Mediterranean-type setting. Both models will be discussed in more detail during the workshop.

Neither of these models has had much impact on development in the northern Negev. That would safely classify them with the other models except for two recent developments. The Migda research results are being tested in a farm-scale project/ sponsored by the Ministry of Agriculture and the Settlement Department. The project was planned with some of the model analyses in hand. If the project gets off the ground, it will also provide an opportunity to test the UNGAR management models. Too much depends on a rather shaky project plagued with an uncertain future.

The multiple-goal GRAPHICAL model has been used in designing a research project on forage shrubs, funded by USAID. Despite doubts as to the value of forage shrub development under local conditions, the model indicated that even a small positive effect of the shrubs on weaning rates, whether because of greater ewe fertility or greater lamb survival, would justify their introduction on purely economic grounds. Consequently, a set of experiments was designed to test this hypothesis.

Lessons

If proof is needed that some people do not learn from experience, then the Israeli experience in livestock modelling could serve. However, this may not be a bad thing, because so often past experience is not much use in a rapidly changing world where so many problems solve themselves, generally by becoming worse, irrelevant or overshadowed by greater problems. In addition, galloping technology inflation constantly fosters the illusion that a whack of megabytes concentrated into a nanosecond will crack at least some of the painful problems that prevent us managing our affairs more rationally. So even if it is futile to learn the lesson, there may be a point in drawing some conclusions that could serve as an ad hoc basis for discussion if not as a guideline for future modelling activities.

Firstly we need to correct the third law of simulation which says that "simulation modelling will go on until the budget runs out." With computer costs plummeting and with in-house minis and in-office micros proliferating, there is, unfortunately, no serious budgetary constraint. So today it seems that simulation modelling goes on until retirement or decease unless the modeller is fired previously or has given up of his own accord. So if the activity is inevitable, let us try to make it as useful and, as interesting as possible. Mainly, this means finding a worthwhile objective that can be attained by a model analysis better or sooner than by other means. The difficulty with this is that partisan interests tend to cloud the issue so that it is often easier to build a model than it is to obtain consensus on an objective that meets these criteria. Nevertheless, there is virtue in trying to formulate good questions for all the glib answers we tend to generate.

Secondly, it would appear that complex methodology is not necessarily the most effective means of solving complex problems. If our experience is any indication, the most powerful and significant models have been the simpler ones. Most of the more complex models seem to get bogged down in their own complexity. However, complexity in itself would not be so heavy a burden if the model was sound conceptually. If this is the case, the wider generality of the model could possibly justify its cumbersome structure. On the other hand, the difficulty with simple models is that they generally require greater intellectual effort. They depend on the ability to perceive the essence of a problem and to pose the significant questions. They usually require more than a modicum of mathematical ability in order to derive their full implications, and they tend to leave one in the lurch when facing specific management problems.

Thirdly, if a model is meant to be used in herd management, it has far more chance of meeting this objective if it is developed in close liaison with the managers or consultants or extension officers who are going to use it. Whether this is because of the obligation that goes with involvement and identification with the model or whether it is because the model is more sharply focused on a real problem is not immaterial, but the chances are that both the motivation to use the model and its usefulness will be enhanced. At least, it makes the definition of the objective a collective responsibility which requires a considerable degree of consensus.

All this has been said before and, at this stage, borders on triviality. Does this mean that such conclusions are misleading? Or that they are valid, but difficult to practise? Or that, in fact, they are being practised to varying degrees; and that the modelling of livestock production is indeed maturing; and that we'll hear about major advances and achievements in the course of this workshop? That would be gratifying and justification enough for the effort involved in organising it. Regarding the Israeli experience, we can point to a significant achievement in the elaboration of theory in grazing dynamics and to some extent in system management. The impact of the management models in practice has been far less impressive. Yet one cannot conclude that all the effort has been in vain because there are signs that the need for these models and an appreciation of their special contribution is growing. How soon system analysis will become routine in management and planning of extensive livestock production is difficult to estimate. There are a number of definitions of a wise man in the Hebrew tradition. One of them asks: "Who is the wise man?" and answers: "He who foresees the course of events." That definition would make any estimate of mine a mere guess. I am sure that the wiser among us will point the way more clearly.

References

Benjamin Y. 1983. A management model of a grassland sheep system under Israeli conditions. M. Phil. thesis, Univ. of Reading, UK.

Goldman A, Seligman N G, Amir S, Drori D and Halevi A. 1977. A nutrition model of dairy cattle and its possible uses. Special Publication No. 98, ARO, Bet Dagan (Hebrew with English summary).

Kahn, Hava E. 1982. The development of a simulation model and its use in the evaluation of cattle production systems. Ph. D. thesis, Univ. of Reading, UK.

Kahn, Have E and Spedding C R W. 1983. A dynamic model for the simulation of cattle herd production systems: I. General description and the effects of simulation techniques on model results. Agric. Systems 12:101–111.

Kahn, Hava E and Spedding C R W. 1984. A dynamic model for the simulation of cattle herd production systems: II. An investigation of the various factors influencing the voluntary intake of dry matter and the use of the model in their validation. Agric. Systems 13:63–82.

Kahn, Hava E and Lehrer A R. 1984. A dynamic model for the simulation of cattle herd production systems: III. Reproductive performance of beef cows. Agric. Systems 13:143–159.

Noy-Meir I. 1975a. Primary and secondary production in sedentary and nomadic grazing systems in the semi-arid region: analysis and modelling. Final research report submitted to the Ford Foundation (Project 7/E-3). Department of Botany, Hebrew Univ., Jerusalem, Israel.

Noy-Meir I. 1975b. Stability of grazing systems: an application of predator prey graphs. J. Ecol. 63:459–481.

Noy-Meir I. 1976. Rotational grazing in a continuously grazing pasture: a simple model. Agric. Systems 1:87–112.

Noy-Meir I. 1978a. Stability in simple grazing models: effects of explicit functions. J. Theor. Biol. 71:347–380.

Noy-Meir I. 1978b. Grazing and production in seasonal pastures: analysis of a simple model. J. Appll. Ecol. 15:809–835.

Seligman N G, Tadmor N H and Dovrat A. 1972. An exercise in the simulation of a semiarid Mediterranean grassland. Bull. Rech. Agron. Gembloux: Volume extraordinaire été a 1'occasion de la semaine d'étude des problèmes Méditerranéens: 138–144.

Seligman N G, Benjamin R W and Eyal E. 1981. Migda Systems (MIGS1). A model for studying management systems of an integrated sheep-wheat farm in the semi-arid zone of Israel. Special publication no. 207. Division of Scientific Publications, Volcani center, ARO, Bet Dagan, Israel.

Seligman N G. Spharim I, Harel Dalia, Eyal E, Benjamin R W and Benjamin Y. 1983a. Economic evaluation of agropastoral systems based on wheat and sheep husbandry applied to the northern Negev of Israel: 1. Effect of area of pasture in relation to area of wheat. Hassadeh 63:877–880; 2. The effect of three-year legume pasture and of cheaper recycled feed on economic feasibility. Hassadeh 63:1126–1128. (Hebrew, with English summary.)

Seligman N G, Spharim I, Eyal E and Benjamin R W. 1983b. Definition of a feasible agriculture technology in terms of multiple socio-economic goals: A graphical method applied to dryland agro-pastoral development in the reviewed northern Negev of Israel. Pamphlet no. 226, Division of Scientific Publications, Volcani center, ARO, Bet Dagan, Israel. (Hebrew with English summary.)

Seligman N G. 1983. Forage shrubs and the economic viability of agropastoral systems in the Mediterranean semiarid zone. AID/CALAR/FOPAR Research Report 2–83, United States–Egypt–Israel Cooperative Arid Lands Agriculture Research Program: Fodder Production and Utilization by Small Ruminants in Arid Regions.

Seligman N G and Spharim I. 1984. Socio-economic feasibility of development in agropastoral systems: a quantitative approach. In: Proceedings of the 2nd International Rangeland Congress, Adelaide, Australia. (in press).

Seligman N G, Spharim I and Benjamin R W. 1984. Nitrogen fertilizer application and income stability in semi-arid pastoral systems. In: Proceedings of the 2nd International Rangeland Congress, Adelaide, Australia. (in press.)

Spharim I and Seligman N G. 1983. Identification and selection of technology for a specific agricultural region: A case study of sheep husbandry and dryland farming in the northern Negev of Israel. Agric. Systems 10:99–125.

Spharim I and Seligman N G. 1985. A graphical method for relating multiple socio-economic goals to R & D objectives in agriculture. Res. Policy. (in press.)

Tadmor N H, Noy-Meir I, Safriel U, Seligman N G, Goldman A, Katznelson J, Eyal E and Benjamin R W. 1977. A simulation model of a semi-arid Mediterranean grazing system. Israel J. Bot. 26:161–171.

Talpaz H, Seligman N G, Goldman A, Sklan D and Horwitz S. 1980. Optimal. trajectory of lactation and nutrition for the dairy cow. In: D Yaron and C Tapiero, eds. Operations research in agriculture and water resources. Proceedings of an International Conference, Jerusalem, 1979. North Holland Publishing Co., Amsterdam. pp. 285–294.

Ungar E D. 1984. Management of agropastoral systems in a semi-arid region. Ph.D. Thesis, Hebrew University of Jerusalem, Israel.

Weitz M and Seligman N G. 1985. BEEFX-Management analysis of a beef herd grazing seasonal Mediterranean range. Research Report, Division of Range and Forage Crops, ARO, Bet Dagan, Israel.

Zaban H. 1981. A study to determine the optimal rainfall land use systems in a semi arid region of Israel. D. Phil. thesis, University of Reading, UK.

Discussion

Question – From a farm manager's point of view, should not one ask what the aims of the model are: does it in fact help the farmer, improve his knowledge or hindsight?

Answer – These are questions that should always be asked, i.e. what are we aiming at? What are we doing it for? Who is our target audience? Is it the scientific community that we want to impress, is it the economic investor whom we want to influence, is it the farmer in the field we want to change or help? Very often, one may find that if we can determine clearly who the target is, many of the other problems are going to fall into place, simply because the audience determines what the preferences, what the priorities are.

Question – Is there any danger, as perhaps there has been in the past, of the model being the objective?

Answer – Very often, if you don't really determine who or what it was being done for, the actual activity itself becomes quite addictive. It is a responsibility of the modeller and of the people around to see that his effort is actually serving some useful purpose. Sometimes people just stop, and that's where it ends; sometimes it is taken up and seems to help other people carry on doing their modelling. Whether they come up with anything better is also often dependent on who is doing it, how intelligently it is being done, and how much luck a person has in striking the happy combination of a problem which is manageable and having the necessary environment and data and people who are able to carry things on. A very good example would be the TAMU model that started off by people taking the initiative in Texas, using a lot of public relations, and getting the model applied on a very large scale in many other parts of the world. The model apparently had a fairly sound fundamental basis, because quite a number of people have used it as a basis for developing their models further, such as the Kahn and ILCA models. There we have an activity that has provided a sort of framework for quite a lot of subsequent modelling activities in the world at large.