Relational databases for the design and construction of maintainable decision support systems in agriculture

Abstract

A general methodology for the design of decision support systems has been tested in an agricultural application, namely the diagnosis of nutrient deficiencies in an annual pasture legume, Trifolium subterraneum. The method models expertise as data (the indivisible objects of the system), information (explicit associations between data items) and knowledge (implicit associations between information and/or data items). To implement the model the information is stored in a relational database and the knowledge defines how further information items may be derived from the current state of this information base. The methodology is illustrated with expertise for diagnosis of nutrient deficiencies. The use of a relational database makes the information component of the system explicit and allows it to be normalised (having no redundancies or ambiguities). The information is readily accessible and easily modified or updated. Such transparency can be extended to the knowledge component, resulting in a very open system that is easily updated or extended. Since the underlying model is general purpose, the system can also serve purposes further than decision support, such as education, training and information. These properties are particularly attractive in an agricultural domain where decision support systems often use large amounts of information, and where researchers and farm advisers relocate regularly whereas projects usually remain fixed.