Water-quality-based BMP design approach and uncertainty analysis for integrated watershed management

This dissertation has developed a water-quality-based approach to design BMP management strategies at the watershed scale. In addition, it has extended existing uncertainty analysis techniques to investigate the propagation of uncertainty inherent in model parameters within a coupled model system as well as to evaluate the effects of uncertainty on model outputs and decision-making processes. The methodology involved first developing a BMP design approach by applying deterministic watershed and water quality models, and then evaluating BMP strategies by considering the effects of uncertainty in the model parameters.; A coupled model system, consisting of a watershed model (HSPF) and a water quality model (CE-QUAL-W2), and a stochastic simulation approach were utilized to design alternative BMP management strategies at the watershed scale. The water-quality-based approach links the in-stream water quality directly to BMP performance at the watershed level, and hence can better guarantee that the developed BMP strategies will comply with water quality requirements.; Uncertainty analysis to investigate the propagation of parameter uncertainty within the coupled model system and to evaluate decision risks of the developed BMP scenarios involves applying four methods in sequence: First, Sensitivity Analysis was used to identify and select the sensitive model parameters; Second, First-Order Error Analysis was applied to quantify and propagate the contribution of each selected variable to the overall variances of model outputs; Third, Monte Carlo Simulation was employed to draw the probability distributions of model outputs with respect to identified parameter uncertainty and to evaluate the risks of BMP scenarios developed from the water-quality-based design approach; and finally, Bootstrap method was adopted to further estimate the decision risks of BMP scenarios with consideration of uncertainty in water quality criteria in addition to model parameter uncertainty.; Uncertainty analysis showed that: (1) of the large number of model parameters, only a few significantly affected the variation of pollutant loads at the watershed outlets and concentrations in the receiving water-body; (2) Uncertainty in parameters related to the limiting factor have great effects on simulated concentrations for a eutrophication problem; (3) significant sources of uncertainty for a non-point pollution problem could differ depending on different management endpoints; (4) First-Order Error Analysis may not be an appropriate method to address the combined effects of uncertainty in a coupled model system on water quality results due to the nonlinearity; (5) Some BMP scenarios showed finite probabilities of violating the target water quality criterion when uncertainty was incorporated into model simulation. Feasible alternatives can be evaluated and ranked based on their risks of exceeding the target water quality criteria, and (6) decisions regarding BMP alternatives based on their relative risks will likely differ depending on whether a deterministic or a probabilistic criteria is applied.