Assessment of uncertainty in optimal watershed management to control nonpoint source pollution from agricultural watersheds

Chetan Maringanti, Purdue University

Abstract

Best management practices (BMPs) provide a viable option, when implemented properly at a farm level, for reduction of nonpoint source (NPS) pollutant loads at a watershed scale. However, the watershed model used to simulate the BMPs is prone to several uncertainties. The important sources of uncertainty are the uncertainty in the estimation of the hydrologic model parameters, uncertainties in the land use and uncertainties in future climate. The main goal of implementing BMPs at a watershed is to place the management practices at a farm level that meets the two objectives of (a) minimization of pollutant loads and (b) minimization of total costs due to the implementation of BMPs. Some of the most common methods for the selection and placement of BMPs in a farm are (a) random selection based on a first come first serve basis, (b) targeting the farm areas that fall within the high contributing areas in the watershed, and (c) optimization method using multi-objective optimization. In this research we applied a multi-objective optimization tool for optimal BMP management and compared the results with several targeting and random placement techniques. Optimization solutions were far better when compared to any targeting and random placement techniques. Uncertainties in the land use change arise from the lack of knowledge on how much the land conversion leads to conversion of forests into croplands or urban areas. This source of uncertainty was studied in this research by developing synthetic land use change scenarios that have varied levels of conversion of land use from forest to croplands (corn or soybean) or forest to urban. It was observed that land use change impacts the water quality loads in the watershed. Climate change is another important source of variability in estimating water quality loads. Downscaled and bias corrected GCM data were used to obtain the variability in the future climate when compared to the historic climate data. This variability was used to modify the historic observed climate data and simulated with the Soil and Water Assessment Tool (SWAT) model to study how the variability in the historic averages would impact the water quality loads and therefore the optimal BMP solutions. Climate change is an important source of uncertainty, and water quality loads predicted by the SWAT model are highly sensitive to variability in climate change. The conclusions from the land use and climate sensitivity analysis was followed by a Monte-Carlo based uncertainty analysis to understand how the SWAT model parameters in combination with the land use and climate variables impact the water quality loads in the watershed. The uncertainty distribution obtained after the Monte-Carlo uncertainty analysis was used to estimate BMP pollution reduction effectiveness using a Latin-Hypercube technique. The variability in the BMP pollution reduction indices due to the uncertainty in the model parameters, land use, and climate change provide uncertainty bands around the BMP optimization solutions. These uncertainty bands around the Pareto-optimal fronts at the end of optimization provide useful information for the decision makers by providing a better understanding of the reductions that can be expected for any particular amount of money invested in the implementation of BMPs and vice versa.

Degree

Ph.D.

Advisors

Chaubey, Purdue University.

Subject Area

Natural Resource Management|Agricultural engineering|Water Resource Management|Environmental engineering

COinS