Albuquerque, NM — Sandia National Laboratories researchers are working with the U.S. Environmental Protection Agency (EPA), University of Cincinnati and Argonne National Laboratory to develop contaminant warning systems that can monitor municipal water systems to determine quickly when and where a contamination occurs.

It’s all part of the EPA’s Threat Ensemble Vulnerability Assessment (TEVA) program to counter threats against water systems. The program uses a computational framework containing a suite of software tools that can simulate threats and identify vulnerabilities in drinking water systems, measure potential public health impacts, and evaluate mitigation and response strategies.

The EPA became particularly concerned about potential water system contamination after the Sept. 11, 2001 attacks on Washington, D.C. and New York.

U.S. water systems consist of large networks of storage tanks, valves, and pipes that transport clean water to customers over vast areas. By the very nature of their design, they provide multiple points for potential contamination — either accidental or intentional.

Sandia is a National Nuclear Security Administration laboratory.

“Our involvement dates back about three years ago when the EPA became aware of some LDRD [internally-funded Laboratory Directed Research and Development program] research we were doing to model threat assessments to water systems,” says Sean McKenna, Sandia project researcher. “We started working with the EPA in March 2003.”

During the ensuing three years, the collaborative team created world-class software to address water security issues. The software can aid in the placement of sensors during the design stage of a contaminant warning system. It can also determine when and where a contamination event happens, track changes, and determine when the event is over.

“Through careful adaptation of classical algorithms, we are able to solve sensor placement problems on networks that are 100 times larger than those previously cited in the water security literature,” says Jon Berry, who works on sensor placement methods for the project. “Our team recognized and exploited mathematical structure that hadn’t been associated with water security before.”

Bill Hart, Sandia project lead, says the software “helped the EPA meet several internal milestones over the past year,” including developing a contaminant incident timeline for the EPA’s WaterSentinel program and working with a large city water utility to determine the best locations for sensor placement. The WaterSentinel Program is being developed in partnership with select cities and laboratories in response to a Homeland Security Presidential Directive that charges the EPA to develop surveillance and monitoring systems to provide early detection of water contamination.

The EPA will test Sandia’s event detection methods later this summer at a large water system.

“These tests [that the EPA will conduct] will assess the event detection methods so that we can better understand how to respond more intelligently to contaminations as they occur,” Hart says.

Sandia is also leveraging this project with another research project funded by the American Water Works Association Research Foundation to develop a sensor simulator that offers a more complete understanding of how contaminant warning systems may ultimately function when operated in water distribution systems. Sandia researchers are developing a software algorithm that mimics the performance of water quality sensors in common use today.

Sensor characteristics such as noise, drift, and sampling frequency are incorporated into a user-friendly software module that enables system designers to assess on-line data signals for event detection that also take into account imperfect sensors and changing water quality baselines that are encountered during routine system operation.

The event detection methods and its sensor simulator have been specifically tailored for use with a variety of affordable, off-the-shelf sensors commonly used by water utilities to monitor water quality.