Structure of agricultural simulators: A philosophical view

Abstract

Model building is as much an art as a science. We attempt in this paper to define modelling terminology and discuss major steps in model building, analysis and synthesis and the sequence in which they are commonly undertaken in agricultural simulation projects. Model building tends to be an iterative process, meaning that at any point in the sequence the model builder may return to a previous step. This illustrates one of the important steps of a successful simulation—the earlier steps are always subject to revision as additional information is obtained. For example, a review of data often causes some reformulation of the problem. This reformulation may involve only a change in problem emphasis, or it may result in complete restructuring. In dealing with complex systems, analysts sometimes find that the first definition of the problem needs later revisions because it included too many (or sometimes too few) factors to represent adequately the alternatives and, therefore, a return to design. Availability of data is always an important consideration and is part of the problem formulation step of model building. Model constants and parameters, as well as initial values for all variables, are derived from these data. In addition, model assumptions are based on available data, giving model predictions some level of credibility. Finally, data must be available for model validation studies. Therefore, as pointed out in this paper, data sources should be located and their adequacy evaluated early in the model building sequence. The programmer or analyst programming the model should size up the task as an initial step. The result is serious if the size or complexity of the job is underestimated and the model will probably never function as economically or as efficiently as it should. Programming language choices and computing systems available are a part of this task. Consideration of model results may not be the end product of the modelling effort. We stress in this paper that the model building process should be an inter-disciplinary effort, so that when the team considers model results, each member has some realisation of their derivation. It is entirely possible at this point that the team decides more detailed or different results are necessary and that the model building process returns to the beginning with new objectives in mind. We attempt to give a detailed and thorough discussion of the model building process as discussed above. However, we hope the process discussed in this paper will be taken only as a philosophical view and should not be treated as a method that automatically produces the creative thinking required to translate a complicated real-world agricultural system into a more compact and usable simulation model.