AWWA WSC67773

A Contaminant Warning System (CWS) seeks to quickly detect contamination within a water distribution system so that action can be taken to minimize consequences. The goal of the U.S. Environmental Protection Agency's (USEPA's) Water Security Initiative (WSi) (formerly WaterSentinel) has been to design, deploy, and evaluate a model CWS. Deployment of the CWS at the first pilot utility in Cincinnati, Ohio occurred in Spring of 2007. Water quality monitoring is one component of the CWS pilot, and inclusion of this component is based on a growing body of data that suggests anomalies in standard water quality (WQ) parameters may provide an indication of contamination. Deviations in one or more WQ parameters from an established base-state can provide early warning of contamination, but only if the anomalies can be picked out of noisy background data. To facilitate this, event detection systems (EDS) are used which rely upon algorithms of various degrees of sophistication to detect anomalous conditions. While such tools have been used in other applications, there is little experience applying them to the detection of anomalies in drinking water systems. The WSi-CWS pilot provides a unique opportunity to collect data necessary for a rigorous evaluation of EDS tools, including data from 17 WQ monitoring stations, each with a suite of 7 WQ sensors. So far, 2 EDS tools have been implemented at the pilot utility, and evaluation of these tools continues with the objectives of improving understanding of EDS tool capabilities, quantifying the performance of EDS tools, and investigating the influence of factors such as site WQ variability on EDS tool performance. In the study described in this paper, data from Greater Cincinnati Water Works was used to evaluate two EDS tools, Canary [Hart, et al, 2007] and H₂O Sentinel [FTI, 2006]. This paper provides some results from the evaluation and discusses some lessons learned from implementing the evaluation. All examples and figures used in this document were taken directly from data and results from the 2007 evaluation. However, location names and details about events are removed to maintain security and confidentiality. Includes 3 references, tables, figures.