Uncertainty analysis for knowledge-based decision aids: Application to PDSS (phosphorus decision support system)

The uncertainty associated with predictions generated by computer decision support systems is often not quantified, First order uncertainty analysis was adapted to quantify the uncertainty in the prediction of phosphorus requirements by the phosphorus decision support system (PDSS). Uncertainties were evaluated on the basis of input values of each variable rather than using the population mean and standard deviation as in standard first order uncertainty analysis. The comparison between first order uncertainty analysis and Monte Carlo simulation showed a good agreement when the coefficient of variation of the buffering coefficient was small. Components of overall uncertainty in PDSS were identified and compared. Crop critical level and buffer coefficient contributed the most uncertainty -a more than variation in extractable P or rate of the P slow reaction. The identification and ranking of the source of variation helps prioritize research to improve recommendation precision. (C) 1997 Elsevier Science Ltd.