Quality of modelling for Integrated Crop Management : Issues for discussion

Models are used in research and extension, to draw together information and to suggest actions. In combination with computers, models constitute frequently used means of information transfer within and among groups of people in agricultural knowledge systems: researchers, farmers, policy makers, and consultants. Assessment of the effectiveness of models as means of communication is topic of a Study Group of the International Organisation for Biological Control, West Palearctic Section. To date, the Study Group has organised workshops together with groups working on greenhouse commodities (van Lenteren, 1996) and orchard management (this volume). Decision support systems (DSS) provide a framework to structure ecological and management information and enhance efficient use of financial and natural resources. Despite these promising traits, the contribution of DSS to practical Integrated Crop Management (ICM) in orchards has been limited. Most DSS have been developed solely based on expertise and empirical information. The drawback of such a purely empirical approach to designing and managing ICM systems is that systems in new commodity groups or new additions to existing systems need to be made by trial and error. The area of validity is limited, but the actual limits are often not clear. Among the causes of failure, lack of clarity on the role of DSS in learning and decision making, as well as organisational problems related to user support and updating the model seem to stand out. The currently prevailing mismatch between demand for and supply of quantitative information is a problem for (applied) research and practice. Within the research community, the wide-spread adoption of systems analysis and process-based modelling as research tools has its problems, too. Process-based models are often presented as new, even if they use ecological concepts that have been used many times before. The application in a specific system may be new, but the concepts may be well known and often applied. The newness and importance of process-based models is hard to assess unless explicit statements are made regarding relationships to existing models: ecological concepts, underlying data, and implementation (programming language, hardware requirements, etc.). Models are developed for specific purposes and environmental conditions, and are tested accordingly. However, lack of specification of the domain of validity may lead to applications for which the model was not intended. Disappointing results an