Assessing Uncertainty in Chlorine Residual Predictions in Drinking Water Distribution System

One common application of network water quality models is chlorine residual prediction. These applications typically use estimates of bulk chlorine decay kinetic parameters derived from bottle test experimental data. While various kinetic models have been developed to predict chlorine residual loss from these bottle tests, and thus in the water distribution system, previous work has not adequately addressed kinetic parameter uncertainty and consequent model prediction uncertainty. In this paper, Monte Carlo simulation is used to estimate chlorine kinetic parameter uncertainty and propagate the resulting uncertainty to network chlorine residual predictions. The consequences of choices regarding the type and quantity of chlorine bottle test experimental data and the type of chlorine kinetic model are also explored. Two kinetic models are investigated for the bulk decay of chlorine residuals: the widely used first-order model, and a second-order model with respect to chlorine and a fictive reactant. Results show that the second-order model can perform better than the first-order model under rechlorination conditions performed in laboratory tests, but the difference in model predictions within the water distribution system depends on residence time, mixing and rechlorination options. Moreover, the parameter estimates for the second-order model can be highly uncertain when using single dose bottle tests data. This uncertainty is reduced when parameters are estimated using data from multi-dose experiments.