Robustness of Designs for Drinking Water Contamination Warning Systems under Uncertain Conditions

Abstract

Contamination warning systems (CWSs) for drinking water distribution systems (WDSs) are used to reduce the potential adverse effects of intentional or accidental WDS contamination. They are designed on the basis of possible contamination events but often address only a narrow range in event conditions. The influence on their performance of conditions different from those assumed in the design generally is not considered. Using results from simulations done with network models for 11 actual WDSs, it is shown here that CWS performance for high-consequence events can degrade substantially (by an order of magnitude) when conditions such as contaminant toxicity and injection time differ from those used in the design. Generally, increasing the number of sensors does not reduce this sensitivity to changed conditions. The significance of uncertain conditions varies substantially among WDSs. As a consequence of performance changes that occur when conditions change, mean-case designs generally outperform worst-case designs when the objective is to minimize worst-case adverse effects over a range of conditions. The results of this work can be used to implement more robust designs for CWSs, while reducing computational requirements.