Testing and validation of an expert system for making liming decisions

• Published in: Nutrient cycling in agroecosystems Vol. 54; no. 1; pp. 9 - 20
• Main Authors: Dierolf, T.S. (Jasa KATOM, Bukittinggi, West Sumatra (Indonesia).), Amien, L.I, Yost, R.S
• Format: Journal Article
• Language: English
• Published: Dordrecht Springer Nature B.V 01.05.1999
• Subjects: ACID SOILS; Acidity; Agricultural production; AIDE A LA DECISION; ALUMINIO; ALUMINIUM; APLICACIONES DEL ORDENADOR; APPLICATION DES ORDINATEURS; Cassava; CHAULAGE; COMPUTER APPLICATIONS; Crops; DECISION SUPPORT; Decision support systems; ENCALADO; Expert systems; Farms; Glycine max; Knowledge bases (artificial intelligence); LIMING; Oryza; Saturated soils; Saturation; Soil lime; SOL ACIDE; SOL TROPICAL; SOPORTE DE DECISION; Soybeans; SUELO ACIDO; SUELO TROPICAL; Toxicity; TROPICAL SOILS; Workers
• ISSN: 1385-1314; 1573-0867
• DOI: 10.1023/A:1009749305556

Computer-based technologies are becoming increasingly important in translating the wealth of agricultural research information into forms that can be used by extension workers. The Acidity Decision Support System (ADSS) is a computer-based expert system that diagnoses and corrects Al-toxicity problems on highly-weathered mineral soils. The objectives of this work were to use published data to evaluate and provide recommendations to improve the ADSS knowledge base. The relative yield-Al saturation curves used by ADSS satisfactorily predict the results of field experiments for maize and soybean, but needs to be improved for rice and cassava. Curves used by ADSS for all four crops should, however, be updated to represent the research data better. Insufficient data were obtained to evaluate other crops satisfactorily. ADSS predicted that more lime would be needed to reduce soil Al saturation than was actually needed. This can be corrected by changing the value of the liming factor (LF) constant in ADSS to better represent the capability of liming materials to neutralize soil Al in the field. ADSS also predicted more response of rice and soybean to lime in on-farm experiments than occurred in practice, although there was a correlation between predicted and actual relative yield for soybean. ADSS should indicate the chances that the predicted relative yield response will not occur because of factors other than liming, under normal farming conditions. The Smyth and Cravo equation provides a reasonable estimate, in lieu of alternatives, of the change in soil Al saturation for up to five years after liming.