Effect of Water Quality Model Uncertainty on the Passaic TMDL and Water Quality Trading Program for Total Phosphorus

Water Quality Model Uncertainty Analysis of a Point-Point Source Phosphorus Trading Program1

“…Specifically, for a given location and state variable, the actual proportion of success (i.e., proportion of observations within the predicted 10th and 90th percentiles) fell within the 95% confidence interval about the expected proportion of success, as detailed in Kardos (2009). These sources can be broadly categorized as parameter and input uncertainty.…”

“…The following areas were outside the scope of the uncertainty analysis, with the exclusion of all but the Wanaque Reservoir model justified by the aforementioned statistical analysis in Kardos (2009): hydrodynamic model uncertainty (including WWTP effluent flows), nonpoint source load model uncertainty, model structure uncertainty, measurement uncertainty, and the uncertainty associated with the proprietary LAWATERS model of the Wanaque Reservoir (Najarian Associates, 2005). LAWATERS uses diverted phosphorus load output from the PWQM, and phosphorus loads from tributaries naturally upstream of the reservoir, to simulate state variables in the Wanaque Reservoir.…”

“…With x = 35, the model was run 50 times. Parameters with skewed distributions utilized the PWQM calibrated values as either the mode or geometric mean, whichever resulted in a more successful comparison of uncertainty estimates to observations, as determined by the statistical approach described in Kardos (2009). The PWQM was calibrated and validated by Omni Environmental Corporation (2007), yielding a fixed set of parameter values that reasonably matched observations of multiple state variables throughout the watershed.…”

“…The PWQM was calibrated and validated by Omni Environmental Corporation (2007), yielding a fixed set of parameter values that reasonably matched observations of multiple state variables throughout the watershed. In the absence of available literature for a given parameter, trials with different distributions were performed until model output uncertainty estimates successfully bounded observations, as determined by the statistical approach described in Kardos (2009). For kinetic parameters with symmetric distributions, the PWQM calibrated values served as mean values.…”

“…This assumption was validated by a statistical comparison that tested how well uncertainty estimates matched observations. Specifically, for a given location and state variable, the actual proportion of success (i.e., proportion of observations within the predicted 10th and 90th percentiles) fell within the 95% confidence interval about the expected proportion of success, as detailed in Kardos (2009).…”