Research Highlights

Threat Ensemble Vulnerability Assessment: CANARY, Event Detection Software

The figure shows water quality data in a distribution system over a period of 12 hours. Near hour 8, a contamination incident has occurred, and the changes in the water quality signals are evident.

Problem statement

Anomalous water quality monitoring results in a drinking water system can indicate the onset of a contamination incident. For this reason, water utilities must be able to rapidly and accurately identify such results. Unfortunately, this is a difficult task because water quality varies significantly from location to location and over time. Software that can interpret water quality data in real time can greatly enhance detection. EPA has partnered with the Department of Energy’s Sandia National Laboratories to develop the CANARY data analysis software to assist water utilities in detecting contamination.

CANARY software is being piloted in five U.S. cites (Cincinnati, c New York, Los Angeles, Philadelphia, and San Francisco) and Singapore, using data sets that are unique to each system. Based on data analysis performed by CANARY, the pilot utilities can better detect when they need to alert their customers to a hazardous level of contamination in their drinking water supply. This in turn, enables them to incorporate new guidelines into their emergency response and consequence management plans.

CANARY Software Information and Availability

CANARY software evaluates standard water quality data (e.g., free chlorine, pH, total organic carbon) over time and uses mathematical and statistical techniques to identify the onset of anomalous water quality incidents. Before using CANARY for the first time, historical utility data must be used to determine the natural variation of these water quality parameters. This allows the water utility to adapt CANARY to work accurately at multiple locations within the water distribution system and helps utility operators to understand the expected false alarm rates associated with CANARY and contamination incident detection.

The CANARY software allows for:

The use of a standard data format for input and output of water quality and operations data
The ability to select different detection algorithms (CANARY contains three different mathematical approaches for analyzing the data)
The ability to select various water utility and location-specific configuration options
An online operations mode and an off-line evaluation mode
The ability to generate data needed to establish performance metrics (e.g., false alarm rates)

CANARY assists water utilities in understanding the significance of large volumes of water quality data. It can automatically review incoming data regarding anomalous conditions and alert the water utility if further action is required. Not only can CANARY detect anomalous conditions resulting from contamination incidents, but it can detect unexpected “normal” events, such as a sensor malfunction or a pipe break.

CANARY is a Windows program that runs on nearly any computer. It is written in the MATLAB® m-code language but also employs some Java programming language. The typical user will not need to know MATLAB®; however, modifications to the source code require knowledge of MATLAB® and/or Java.

The CANARY software is not intended to replace commercially available software but to supplement existing software and motivate commercial development of similar products. It is hoped that the release of CANARY will also motivate additional research and development in this field.

Future CANARY Updates

Several research directions are currently being pursued to improve CANARY.

An automated configuring approach within a water distribution system is being developed. This approach will use a set of “normal” historical data from a water distribution system and automatically find a set of “good” parameters to be used in the configuration process. Because event detection software has an inherent trade-off between detection sensitivity and the false alarm rate, each user must eventually decide what the best parameters will be for a specific use of CANARY.
Researchers are investigating different ways to interface with water utilities’ Supervisory Control and Data Acquisition (SCADA) systems to improve the performance of CANARY. SCADA refers to a large-scale, distributed measurement and control system that controls components such as pumps and valves, and records data from sensors.
Researchers are pursuing methods to use information collected simultaneously from multiple locations (rather than each location separately) to increase the sensitivity of detecting contamination incidents.
By incorporating information from several sensor stations, along with information about hydraulic connectivity, CANARY may be able to more accurately and rapidly detect contamination incidents.
In July, 2010, EPA and DOE researchers were awarded R&D Magazine’s “R&D 100” for developing CANARY. The “R&D 100” awards recognize the top high-technology products of the year.
As a free software tool, CANARY is available worldwide to drinking water utilities striving to provide safe water to their customers. The software has been accessed by more than 600 users in 15 countries.

Software Availability

To find out more about the CANARY software and documentation, please contact Regan Murray, (513) 569-7031.