Expert System for Fertilization Management of Rice

Abstract

A computer-based decision support system called RiceFertility has been developed to provide information and recommendations on efficient utilization of fertilizer for the production of flooded rice in Arkansas. Conventional information sources regarding timing and rate of fertilizer applications were consolidated in developing a computer-based tool that generated recommendations quickly for many situations. The text of each recommendation was formulated to be sensitive to the tactical context of the individual problem scenario. The inputs to the expert system included cultural system, rice cultivar, rice growth stage, flood status, soil texture, and soil pH. Appropriate recommendations were generated for rates of early nitrogen (N), maximum tillering N, midseason N, and other soil fertility problems involving salinity, liming, phosphorus, potassium, zinc, and/or sulfur. The system logic was successfully validated in 29 of 31 sample scenarios tested. The set of sample scenarios was defined using the Arkansas Rice Research Verification Trials as a balanced source of fertilizer decisions made in growers’ fields. A method of classifying the results of the validation testing was useful in evaluating the software.