Types, purposes and users of models
Problems related to intake
Benjamin It might be advisable to call upon different disciplines (researcher, manager, planner) in determining the use of models. It is obvious that models have a different meaning, purpose and use for each of these, and that they can be separated in some way.
Few complex models have been used by managers. Simple models have been of more use: in these, some day-to-day management-decisions are reduced to a simple yes or no answer. Another type of model is that with a fundamental approach, e.g. Ketelaars' approach towards intake based on energy and protein ratios. These can be extremely useful to farmers, although energy and crude protein do not mean very much to them. What they really want to know is what they get with small inputs of supplements.
Seligman Benjamin raised two good points. One is the complexity of the models, and the other was the question of how to deal with quality and intake. The latter will be discussed in the second part of this session. Regarding the first, I would like to say that we should be careful with terms like complex and simple. Many of the 'complex' models are conceptually quite simple, while some of the 'simple' models are conceptually very sophisticated.
The usefulness. of a model depends on the system, on the way the model has been conceptualised and on the way in which one goes about the modelling process: some complex models are well conceptualised and are very useful, while some of the simple ones do not mean much. It is important that the final answer from the model should be in a form which can be easily understood by the people who will apply the results.
Lambourne I think that Benjamin has brought up a very important point that is of considerable interest to ILCA. The ILCA model is a modification of the TAMU model. I am beginning to wonder if the ILCA model is not a sort of hybrid, because ILCA's requirements are likely to diverge from those of TAMU. One direction is the use of decision-type management models and practical models. These do not give a great deal of resolution, but they do give an idea of the actual output of a system: figures for herd dynamics and production which give a broad indication of the productivity of different systems. ILCA also needs models that go into greater detail of the physiological basis of plant growth or the reproductive performance of the animals. I do not see that one model, e.g. the ILCA model, can serve both purposes.
Spharim I would like to raise a related subject. It was mentioned that the ILCA model, which is applicable to the conditions in Botswana, does not fit well in a different environment. If the biological basis of the system was well understood, it would be possible to find a factor within the model that would be widely applicable to different environments. Furthermore, if the mechanism was understood, less data would have to be collected in a given place, but the model would still be applicable to that environment.
Upton Possibly one could distinguish between models by the degree of their mathematical complexity. Conversely, complexity can be determined by the number of relatively simple relationships that you are trying to investigate simultaneously. A common notion is that there is a particular constraint that limits the achievement of the objective. If you can pinpoint that constraint, your efforts can be focused on overcoming it.
Seligman That is true. When you want to find the limiting factor in a particular situation, you have to investigate the situation itself. A comprehensive model will define lots of processes, but not necessarily those that deal with the limiting factors. The problem in modelling production systems is to combine these two approaches: the identification of the limiting factors and the elaboration of the model so that it expresses them correctly.
de Ridder I want to go back to the choice of the type of model. The matter is not of choosing a particular type of model, but to choose the one best suited to the defined problem. Starting with a general, low-resolution model, you can try to determine the possibilities for improving production from a system. Much of the input for the model will be data collected from the system itself, but some guesses will have to be made about how the system will react to proposed interventions. A biological simulation model is needed to quantify these guesses, e.g. the effect of increased forage availability on overall production of the system.
Lambourne I agree. One of the justifications for the scientific type of model is that they produce reliable units which can then be used to assess research and development priorities or to assess management decisions in the decision-type models.
Orshan One has to distinguish between problem-oriented and system-simulation models. The first is a deducive and the latter an indicative approach. There is no advantage to using one alone. Both have to be used, but one must distinguish between aims and techniquesthat is the point.
van Keulen In Wageningen we have been modelling biological systems for many years, with the aim of gaining a basic understanding of the biological systems. Some of these models go from a plant to an organ or even cellular level, while others go in the opposite direction, from the plant to the whole canopy level. With both types of model dissatisfaction creeps in. One begins to wonder how the models are to be used. In this there are two approaches that can be adapted. First, you can sit and wait until someone comes along with a specific problem which needs to be solved. This will probably not work, since most people are not very enthusiastic about models. The second approach is to ask yourself how you can use the knowledge you have gained. An example of this approach has been developed in the Netherlands. The same people who developed the detailed crop growth and population dynamics models, with a view to increasing insight into these systems, developed simplified equations from these models which could be of direct use to the farmer, e.g. rules on whether or not to spray a wheat crop.
These rules have been tried out by farmers for several years now. Last year there were 9 000 fields of wheat that were being served by this advisory service.
In our experience, one has to develop fairly detailed and fundamental models, and then bring them down to a level at which they can be used to solve practical problems.
Seligman The success of the Wageningen group in modelling is clear, but I was under the impression that the amount of modelling used in the advisory services is almost nil.
van Keulen Both the population dynamics model and the crop growth model have been reduced to a few lines of FORTRAN dealing with factors relevant to the development of a particular disease on a particular crop at a particular physiological growth stage. This could not have been done unless the detailed models had been developed first.
Ungar I see that some important methodological points have been touched upon. I think that linguistics is important. The problem defines the way in which the reality is conceptualised.
The language that is used depends upon one's outlook. Thus the concept of relative growth rate in the ARID crop model has no meaning, because ARID crop is not concerned with grazing-systems dynamics, but grazing-systems dynamics cannot be explained without that concept. The problem is that model builders do not have a single level on which they wish to provide an explanation, but many. They try to explain not only foreign income, but changes in animal liveweight, herbage growth and moisture in the soil. They are forced to take the last common denominator, and end up explaining soil moisture but not explaining farm income, only simulating it. In order to develop explanatory power, we need appropriate concepts and language. That is why Noy-Meir's models, which used the correct language and concepts to deal with the problem in hand, are so successful. A model that. is more detailed might be more precise and more realistic, but it would lose the explanatory power of the simpler conception.
Upton In a sense the mandate of ILCA is to develop new technologies. But new technologies developed by research institutes need testing through expensive field trials and extensive pilot projects before they can be presented to the farmers. Another application of modelling is to speed up this testing and substitute for field work. It is a means to test a technology, but can never be a complete substitute for field trials.
Spharim There is no difference between the wind tunnel that an engineer uses to check wings and the models that we are talking of now. Even in a wind tunnel an engineer only simulates the real world. The final test of the wing is in the air, but a good engineer will know what is being tested in the wind tunnel and what is not.
Ungar There is a temporal as well as a philosophical sequence to be followed in modelling. Do we need a good scientific animal model before we can build a managerial-type model? Must we have ARID CROP before we can model the management of lambs? Or can we approach the problem from a different point of view?
van Keulen If you are interested in the potential yield of a wheat crop in a Mediterranean environment with 250 mm rainfall, you can cut out a lot of detail, use fairly simple descriptions, and care up with results that are extremely relevant to the given situation. But you should also realise that the model is descriptive rather than explanatory.
de Leeuw So the question is, are we just going to use the observations from the Borana system or do we need to know the details of the competitive elements, e.g. between milk offtake and calf growth?
van Keulen When you are interested only in what will happen in that particular region or system, then you can formulate a site-specific hypothesis, either on the back of an envelope or with a computer. That hypothesis can be used in analysing what will happen if you change certain elements in that system. But if you want to be able to predict what will happen elsewhere on the basis of these data, then you need something a lot more sophisticated.
de Leeuw Let me pose another question. We know that milk is a limiting resource during the dry season. I could modify the system by feeding certain proportions of urea. What I want to know is whether this is economically feasible in the long- or short-term. I want to know the effects of this intervention on the herd development and on overall productivity, both in good and bad years. Do I need the ILCA model to do this, or can I calculate this on the back of an envelope?
van Keulen The relevant question is, can you describe the effect of urea supplementation in order to predict its effects on animal performance over a number of years? If you can, then there is no harm in using the ILCA model. But if you cannot quantify the effects of urea supplementation then you would be fooling yourself by using the model, because the results would be spurious.
Ketelaars I think that, with regard to changes in nutritional conditions, the ILCA model is not an explanatory one, and, as far as I know, there are no explanatory models that can judge the possible effects of such changes. Thus, I think that there are many possible interventions the effects of which cannot be predicted, and these should be investigated in field trials.
Seligman We will leave this topic, and go on to the problems of intake, which is central in interfacing primary and secondary production models.
Lambourne Again, it seems to me that we can probably skip the interface stage altogether, and simply relate animal performance to characteristics of the resource which we are using. We do not want to predict pasture intake, we want to predict animal performance, and to relate this to characteristics of the area. I think that more needs to be done on how to incorporate the ideas of Ketelaars and others and to assess their importance, but we have to do this at a scientific level. I do not see why that should hold up the practical use of models that simply relate resources to animal performance.
de Leeuw If you are asking these questions in relation to the ILCA model, we can put in a number of reasonably realistic limitations to intake with slight modification to the model. If we want to use the ILCA model for all purposes, we have to look at the intake side of the model.
van Keulen In Holland we are confronting problems that mean that we cannot leave the intake side to experimental and theoretical control. We have pastures that are fertilized with high levels of nitrogen: dairy farmers want their cows to produce 4050 litres of milk per day, while the government does not want the ground water and the soil to be polluted, and the environmentalists do not want acid rain. In that situation we have to describe in clear, quantitative terms what the properties of the forage offered to the animals should be in order to meet all these requirements. What little experience we have in Africa indicates that the performance of animals in extensive systems is limited by the fact that they cannot eat enough. If that is the case, then one would like to know whether the limitation is the amount or the quality of the forage available.
Lambourne Am I right in thinking that this is an important field, particularly at the scientific level, and that there are a number of things that we do not understand about it? This probably calls for more careful research, and may be built into a model in the long run. In the meantime, we need to do research to understand the processes of intake and digestion and the factors that are involved.
van Keulen That may be true to some extent, but you need that information in the short-term. For example, with the data that ILCA is collecting in their country programmes, you need to know if the poor performance of those herds is due to a problem in the interface between the feed resources and the animal or if the animals are sick. There is little you can do in the way of giving advice or making recommendations until this is clarified.
Wagenaar I do not agree that models should not be used at present. They should be used as much as possible, keeping in mind that there are gaps in the knowledge on which they are based. These gaps are not a reason to avoid using the model. Meanwhile work should continue to fill in these gaps in the knowledge. If the knowledge is available, it should be incorporated into the model. For instance, Ketelaars' approach could be introduced into the model without knowing much about water intake or mineral requirements. .
van Keulen That is true, but individual modules of the model have to be tested just as in the primary production situation. At a certain stage in our model, the prediction of dry-matter accumulation did not agree with measured data. It is a long way from C02 molecules in the air to the grain at the top of a wheat crop. A deliberate decision was made to change the boundary of the system in this model and to first make sure that the part that models the pathway from the CO2 molecule to the glucose molecule in the plant could be tested. This led to the development of a method for measuring carbon dioxide exchange in the field using a mobile laboratory. On the basis of this, I would suggest that you take out the intake module of the ILCA model and try to validate it on the basis of any data that are available.
