A multi-objective optimization tool for the selection and placement of BMPs for pesticide control

Abstract. Pesticides (particularly atrazine used in corn fields) are the foremost source of water contamination in many of the water bodies in Midwestern corn belt, exceeding the 3 ppb MCL established by the U.S. EPA for drinking water. Best management practices (BMPs), such as buffer strips and land management practices, have been proven to effectively reduce the pesticide pollution loads from agricultural areas. However, selection and placement of BMPs in watersheds to achieve an ecologically effective and economically feasible solution is a daunting task. BMP placement decisions under such complex conditions require a multi-objective optimization algorithm that would search for the best possible solution that satisfies the given watershed management objectives. Genetic algorithms (GA) have been the most popular optimization algorithms for the BMP selection and placement problem. Most optimization models also had a dynamic linkage with the water quality model, which increased the computation time considerably thus restricting them to apply models on field scale or relatively smaller (11 or 14 digit HUC) watersheds. However, most previous works have considered the two objectives individually during the optimization process by introducing a constraint on the other objective, therefore decreasing the degree of freedom to find the solution. In this study, the optimization for atrazine reduction is performed by considering the two objectives simultaneously using a multi-objective genetic algorithm (NSGA-II). The limitation with the dynamic linkage with a distributed parameter watershed model was overcome through the utilization of a BMP tool, a database that stores the pollution reduction and cost information of different BMPs under consideration. The model was used for the selection and placement of BMPs in Wildcat Creek Watershed (located in Indiana, for atrazine reduction. The most ecologically effective solution from the model had an annual atrazine concentration reduction of 30%, from the baseline with a BMP implementation cost of $18 million. The pareto-optimal fronts generated between the two optimized objective functions can be used to achieve desired water quality goals with minimum BMP implementation cost for the watershed.