Evaluating the Uncertainty in Water Quality Predictions - A Case Study

A method for assessing model result uncertainty is presented and applied to a case where a paper mill wastewater is discharged into an estuary in the Southeastern U.S.. Model result uncertainty was quantified by incorporating the uncertainty analysis into model calibration. The two-dimensional, laterally averaged model CE-QUAL-W2 was used to predict water quality conditions. The water quality model was calibrated against field measurements of longitudinal and vertical variations in salinity, dissolved oxygen, and biochemical oxygen demand (BOD) concentrations. A quantitative, multi-constituent criteria for acceptable calibration was used to identify plausible parameter sets. A collection of plausible parameter sets was then identified, and used to assess the uncertainty in dissolved oxygen prediction, and the uncertainty in predicted system response to a reduction in organic matter loading. A search procedure was also developed to minimize the calibration criteria statistic and to assess the range of model predictions. Plausible parameter sets differed widely in their parameter values, and they produced widely different dissolved oxygen concentration predictions. The system response to reduced loading, however, was found to be very similar between the plausible parameter sets.