Prediction and design problems in crop management
Résumé
Crop management is the way a farmer, being aware of environmental constraints, controls installation and development of a crop, then operates exportation of products, in order to reach given objectives. Crop management encounters 3 main difficulties. First, intensive cropping is no longer profitable and sustainable anywhere. As a result, the farmer has to master a variety of management strategies. Second, a cultivation operation may have irreversible consequences. Thus, the farmer must consider all the cropping operations comprehensively. Third, climatic uncertainty prevents the farmer from fully mastering interaction between succeeding cropping operations. This raises the need for specific tools to deal with uncertainty. We first present a systemic point of view on crop management, in which a production system is analysed as a couple of sub-systems : a biotechnical system and a decision system. Then, we express crop management as the resolution of prediction and design problems on the decision sub-system. Two artificial intelligence-based software applications will be outlined in this paper. The first is a simulation model of a grassland fields manager in the French Pyrenean Mountains. The goal is to make a farmer use this predictive tool interactively in order for him to assess his own decision process and ultimately to improve upon it. The second application solves a design problem : "what is the sequence of cultivation operations to apply on a wheat crop so that the farmer attains his production objective with respect to certain organizational and environmental constraints ?" We observe that these two applications share the same software architecture, although underlying concepts and basic algorithms are quite different.