An approach to evaluate BMP optimization and uncertainty in water quality improvement using Bayesian analysis and BMP Tool

A multi-objective genetic algorithm (NSGA-II) in combination with a watershed model (Soil and Water Assessment Tool) is used in an optimization framework for making the Best Management Practices (BMP) selection and placement decisions to reduce the NPS pollutants and the net cost for implementation of BMPs. Shuffled complex evolutionary metropolis uncertainty analysis (SCEM-UA) method will be used to quantify the uncertainty of the BMP selection and placement tool. The sources of input uncertainty for the tool include the uncertainties in the estimation of economic costs for the implementation of BMPs, and input SWAT model predictions at HRU level. The SWAT model predictions are in turn influenced by the model parameters and the input climate forcing such as precipitation and temperature which in turn are affected due to the changing climate, and the changing land use in the watershed. The optimization tool is also influenced by the operational parameters of the genetic algorithm. The SCEM-UA method will be initiated using a uniform distribution for the range of the model parameters and the input sources of uncertainty to estimate the posterior probability distribution of the model response variables. This methodology will be applied to estimate the uncertainty in the BMP selection and placement in Wildcat Creek Watershed located in northcentral Indiana. Nitrogen, phosphorus, sediment, and pesticide are the various NPS pollutants that will be reduced through implementation of BMPs in the watershed. The uncertainty bounds around the Pareto-optimal fronts after the optimization will provide the watershed management groups a clear insight on how the desired water quality goals could be realistically met for the least amount of money that is available for BMP implementation in the watershed.