Uncertainty based analysis of the impact of watershed phosphorus load on reservoir phosphorus concentration

- Uncertainty is analyzed in simulation of P concentration/load of reservoir and watershed using SWAT model.
- A cumulative weighted probability distribution based on Monte Carlo is used.
- FAST method is used to select the effective factors in model outputs variability.
- The BMP implementation in a watershed results in less uncertainty in simulation.
- Incorporating uncertainty in determining the margin of safety in TMDL estimation.

SummaryIn many regions of the world that depend on surface reservoirs as a source of water supply, eutrophication is a major water quality problem. Developing simulation models to evaluate the impact of watershed nutrient loads on the reservoir's water quality is an essential step in eutrophication management. In this regard, analysis of model uncertainty gives an opportunity to assess the reliability and the margin of safety of the model predictions for Total Maximum Daily Load (TMDL) from the watershed nutrient load. In this study, a computational procedure has been proposed for the analysis of the model uncertainties in simulation of watershed phosphorous load and reservoir phosphorous concentration. Data from the Aharchai watershed which is located upstream of the Satarkhan reservoir in the northwestern part of Iran, is used as the study area to test the effectiveness of the proposed methodology. The Soil and Water Assessment Tools (SWAT) is utilized for assessment of watershed phosphorus load as the main agent resulting in the reservoir eutrophication in the region. The most effective parameters in model performance are identified by a global sensitivity analysis technique named modified Fourier Amplitude Sensitivity Test (FAST) which can incorporate parameter interdependencies. The Generalized Likelihood Uncertainty Estimation (GLUE) technique is also applied to set up behavioral ranges of the parameters that are relevant to the actual observations. Finally, the cumulative weighted-likelihood distribution functions (CWLDF) are derived for outputs of the SWAT. They are used jointly for estimation of results uncertainty limits using the Copula method. To assess the effectiveness of applying Best Management Practices (BMPs) in the watershed, two scenarios of with and without BMPs application are tested. The results showed the effectiveness of the proposed model in uncertainty estimation of watershed phosphorus load and reservoir phosphorus concentration as well as the effectiveness of BMPs in reducing P loads from the watershed.