DO Model Uncertainty with Correlated Inputs

The effect of correlation among the input parameters and variables on the output uncertainty of the Streeter‐Phelps water‐quality model is examined. Three uncertainty analysis techniques are used: sensitivity analysis, first‐order error analysis, and Monte Carlo simulation. A modified version of the Streeter‐Phelps model that includes nitrification, net algal oxygen production, and sediment oxygen demand is used. Analyses are performed for a wide variety of simulated stream‐flow conditions. Results show that the standard deviation of the predicted dissolved oxygen deficit (DOD) with correlated inputs potentially can be 20‐40% larger than with independent inputs. Under conditions of moderate to high velocity, the reaeration and bio‐oxidation coefficients are the dominant contributors to DOD uncertainty, while net oxygen production from algal activity and sediment oxygen demand are the major factors at low velocity. The largest effect of input correlation on DOD occurs in the vicinity of the sag point. Uncertainty results from first‐order analysis differ by at most 10% from those of a Monte Carlo simulation for both correlated and independent inputs.