Evaluating best management practices for agricultural watersheds using probabilistic models
This research describes an approach to organize results from a non-point source (NPS) pollution model into a framework for analysis and decision making. A review of NPS models and a survey of techniques used for constructing decision-support software systems is given. The NPS model selected for this investigation was the Guelph model for evaluating effects of Agricultural Management systems on Erosion and Sedimentation (GAMES). The organizational framework consists of a graphical user interface that interacts with a probabilistic model. The probability model integrates probability distributions of important variables and relationships between variables from model simulations. The graphical probability model consists of a network of nodes that represent individual parameters or variables for fields in a watershed. Directed links indicate relationships between variables. The complete graph thus has directed links that follow the drainage network of the watershed. The relationships (or links) between the nodes are quantified by way of data derived by Monte Carlo simulation of the GAMES model and by deterministic functions as specified in GAMES. Probability models were developed for two Southern Ontario watersheds to demonstrate how the system can be used for targeting best management practices. With this technique, it is possible to select any configuration of management practices in the watershed in order to obtain estimates of erosion rates and sediment yield without the need to return to the original simulation model.