Using Monte-Carlo simulations and Bayesian Networks to quantify and demonstrate the impact of fertiliser best management practices

Nutrient exports from agriculture contribute to eutrophication of rivers and lakes. In many jurisdictions Best Management Practices (BMP's) are the cornerstone of mitigation efforts. In this paper we examine the use of Monte-Carlo simulations to combine fertiliser distribution, grazing and runoff data, and regression equations developed from field-scale monitoring, to estimate the maximal effect of fertiliser BMP's on phosphorus (P) exports. The simulation data are then compared with a Bayesian Network that can be used to quickly evaluate the effects of different management scenarios on P exports and communicate those results to landholders. Both techniques demonstrate that for systems similar to those for which the regression equations were derived, improved fertiliser management is unlikely to have a major impact on Total P (TP) exports (i.e. <10%). While the contribution of fertiliser to TP exports in a general sense is relatively small this study suggests that aberrant behaviour (i.e. fertiliser application immediately preceding rainfall runoff) can dramatically increase P exports. The major factor affecting TP exports appears to be the systematic or background P which includes native P and P from previously applied amendments. For communicating the effects of different management scenarios to landholders, Bayesian Networks are shown to be generally superior to Monte-Carlo techniques. However, the study suggests care is needed in selecting the states for the Bayesian Networks and demonstrates that at the extremes, the discretisation required by Bayesian Network software can produce misleading results. (C) 2011 Elsevier Ltd. All rights reserved.