To improve management of resources, such as agricultural crops or forests, scientists attempt to analyse the resource systems and to predict the consequences or outcomes of interventions. They construct models of interactions of components of the systems, drawing on knowledge and experience. In agriculture, five types of models have become common — empirical, crop–weather, crop–growth, crop–system, and crop–process. The models aim mainly to predict crop yields when a series of actions are taken. They differ markedly in complexity, from a simple regression to a series of mechanistic relations aimed at simulating the crop system. The uses to which a model is to be put, and by whom it will be used, are major determinants of the nature of the model so modellers must work with the potential users. In fact, modelling is an exercise in human relations as much as in science. All things being equal, the simpler the model is that meets the objectives of the users, the better are the chances of its being used.

Concerns are frequently expressed within the resource management community that many of the tools or models being developed to help people manage resources aren't being used. Less clear are the reasons, but a look at the contrasts between resource management and resource science provides some clues. The disciplinary organization of scientific knowledge, compared with the general nature of management problems, contributes to the gap between science and management. Many of the difficulties that resource scientists are experiencing are shared by management scientists attempting to provide problem-solving support to managers. In fact, management scientists who are developing tools for use by managers are sometimes said to be doing research on "toy" problems because of their tendency to oversimplify the "messes" or complex systems with which managers are trying to deal. To derive tools of use in solving real problems, some resource scientists will have to be willing to study and work with resource managers in the context where problems are faced.

Two research programs in support of the wheat industry provide insights into mechanisms that can be used in research planning and in technology transfer. Modelling was pivotal in both: in Australia since 1973 in efforts against rice weevil, Sitophilus oryzae L., infestations of stored wheat and in Canada since 1970 against grasshopper infestations of wheat fields. The research studies, described in published papers, were designed by the scientists within broad research objectives defined by the employing organizations. A mixture of technical and human considerations determined the direction of the research: the talents of scientific staff combined to capitalize on fortuitous events or situations. Serendipity improved the response to uncertainty in both research and industrial environments. The nature and amount of effort needed to get scientific insights applied varied with the specifics of the scientific insight. A general analysis of technology transfer suggests that, in some cases, application requires the involvement of a variety of nonscientific specialists and that research managers are well positioned to coordinate the efforts. Taken together, the conclusions show that satisfying a need or solving a problem is a task within which research is only one part. Success requires effective communication and compromises among very different people.

Fusarium attacking the roots of Douglas-firs in a forestry nursery was the focus of a computer simulation model that was tested against several years of nursery disease observations. Soil temperature, seedling growth, disease spread, and inoculum density and distribution were studied and their effects on measures to control Fusarium root rot were simulated. Interactions of these factors in space and time were examined from seedling germination to end of the growing season. The results from the model indicated that low soil temperature at the beginning of the growing season, rapid seedling growth, and reduction of disease inoculum levels in the upper soil horizons were the optimum prescription for disease reduction.

In their simplest form, crop models consist of regressions relating factors such as yield to environmental variables. Explanatory models also exist, with much physiological detail and responsiveness to meteorological and soil factors. Summary models can be derived from such large models, including the influence of pests. Of these major types of models, regression models are increasingly becoming widely used to optimize and integrate the many facets of cereal management in Western Europe. This fact suggests that crop manipulation is evolving toward optimal production management.

A key element of a project designed for integrated pest management in alfalfa fields in New York State is a comprehensive simulation model. The model was developed as a guide for growers to determine when the costs of control measures against alfalfa weevil would be worthwhile. The model incorporates crop and insect submodels. Optimization techniques were used to identify the more cost effective of the possible management strategies. The project has been an unqualified success from a research standpoint but the results are not widely implemented by growers. In this respect, the project has suffered from institutional problems at the research level, lack of confidence and commitment in extension, and skepticism among growers. To rectify this situation, project staff recently adopted new approaches for implementing the program.