Evaluation of Uncertainties in the Simulation of Watershed Nutrient Load: A Case Study

Estimation of watershed nutrient loads into the reservoir through a simulation model is a vital step in eutrophication management of reservoirs. The analysis of uncertainty associated with the utility of simulation models is an important consideration in the development of Total Maximum Daily Load plans in the control of watershed nutrient loading. When using models during the development of the TMDL, either to predict loadings or to simulate water quality, managers should address the inherent uncertainty in the predictions. In this paper, a Monte Carlo based probabilistic approach namely Generalized Likelihood Uncertainty Estimation (GLUE) is utilized to develop a computational procedure for analysis of the model uncertainties on prediction of phosphorous loads of a distributed watershed model, Soil and Water Assessment Tools (SWAT). In the GLUE method a cumulative likelihood for nutrient outputs of the SWAT model is generated. Quantile analysis of the cumulative likelihoods yields expected nutrient loads as well as their corresponding uncertainty limits. Uncertainty bounds could be used to obtain a Margin of Safety (MOS) for such evaluations in the TMDL formula. The analysis is demonstrated for the Aharchai watershed in the North - Western part of Iran.