Generating Different Scenarios of BMP Designs in a Watershed Scale by Combining NSGA-II with SWAT

A multi-objective optimization technique, Non-dominated Sorting Genetic Algorithm (NSGA-II), is coupled with a semi-distributed hydrological model, Soil and Water Assessment Tool (SWAT) with the objective of generating a set of non-dominated solutions. The model is formulated to optimize two conflicting objectives: the cost of structural Best Management Practices (BMPs) and the treatment goal. The cost of the BMPs is the function of type, size, and location of BMPs; whereas, the treatment goal considered is the reduction in annual sediment load the watershed outlet. Structural BMPs included in the modeling include detention ponds, infiltration type detention ponds, parallel terraces, filter strips, and grade stabilization structures which are applicable in agricultural watersheds. The model is demonstrated on Silver Creek watershed, a sub watershed of Lower Kaskaskia in Illinois. SWAT delineates watershed into 159 subbasins and are considered as potential locations of BMPs. The model came up with a Pareto front, which comprises a set of non-dominated solutions, a set of least cost solutions for different level of annual sediment load reduction.