211831 Statistical Models and Water Quality Event Detection

Tuesday, November 10, 2009: 12:50 PM

Sean A. McKenna, PhD , Sandia National Laboratories, Albuquerque, NM
David B. Hart , Sandia National Laboratories, Albuquerque, NM
Mark W. Koch , Dept. of Sensor Explotation Applications, Sandia National Laboratories, Albuquerque, NM
Eric D. Vugrin , Sandia National Laboratories, Albuquerque, NM
Event detection systems (EDS) provide the critical link between continuous online measurement of water quality within a water distribution network and a Contamination Warning System (CWS) that integrates multiple data streams for public health protection. Recent development of the CANARY EDS has relied heavily on statistical models for event detection in environments with high background noise. Online signal filtering and residual classification form the basis of the event detection system. Filtering residuals are classified and combined into a continuous probability of event value using a binomial failure model. False positive reduction has been complicated by water quality changes due to changes in network hydraulic operations. Pattern recognition through trajectory clustering of time series data in multivariate space provides a pattern library. On-line access to this library during event detection is shown to reduce false positives. Further suppression of false alarms can be achieved through integration of the results of multiple EDS tools running independently into a single distributed detection result. All EDS alarms (true and false) are considered to be the result of a spatial-temporal point process and scan statistics are used to identify significant clusters of points within the spatial-temporal domain indicative of true events within the false alarm clutter. Examples of integrating complimentary data streams from a CWS with EDS results are discussed.

Sandia is a multiprogram laboratory operated by Sandia Corporation, a Lockheed Martin Company, for the United States Department of Energy's National Nuclear Security Administration under contract DE-AC04-94AL85000

Learning Objectives:
1. Identify the required comoponents of an operational event detection system. 2. Describe obstacles and statistical solutions for robust detection algorithms.

Keywords: Water Quality, Data/Surveillance

Presenting author's disclosure statement:

Qualified on the content I am responsible for because: I have performed the related research at Sandia Labs under an EPA-sponsored program for several years.
Any relevant financial relationships? No

I agree to comply with the American Public Health Association Conflict of Interest and Commercial Support Guidelines, and to disclose to the participants any off-label or experimental uses of a commercial product or service discussed in my presentation.