Multi-objective modelling and decision support using a Bayesian network approximation to a non-point source pollution model

This paper illustrates a methodology to create a multi-objective modelling system using Bayesian probability networks to emulate the behaviour of an environmental model that was originally intended for the purpose of analyzing a problem – non-point source pollution in our example. Modelling systems frequently pertain to a single domain (physical or chemical process modelling, hydrology or combinations) to simulate a process in nature such as pollution transport or the production of food or manufactured goods. Economic or other effects are treated separately, or handled in a non-interactive manner. Side-effects of agro-industrial processes, or gains/losses from production enterprises, are generally modelled separately without the ability to examine trade-offs or alternatives. Multi-objective modelling attempts to work in more than one problem domain through decision theoretical principles. Such treatments are designed to couple production and waste systems, to quantify the economic cost of remediation. This model will demonstrate such an application, from the data acquisition, model calibration to the hypothesis testing, for a non-point source pollution model. This will be combined with a simplified net revenue model based on crop rotations typically found in Southern Ontario, Canada, using realistic economic data obtained from agricultural operations similar to those found in this region. We will demonstrate that multi-year analyses are possible with such a system.