Effective Use of Real Time Water Quality Data in an Early Warning System – Data Management and Analysis Needs
Oral Presentation
Prepared by T. Bradley1, T. Bartrand2, R. Kopansky3, J. Rosen4
1 - Philadelphia Water - Bureau of Laboratory Services, need, need, need, need, United States
2 - Corona Environmental Consulting, 150 Monument Road, Bala Cynwyd, PA, 19004, United States
3 - Philadelphia Water, , , United States
4 - Corona Environmental Consulting, , , United States
Contact Information: [email protected]; 215-685-1460
ABSTRACT
Online sensor data, whether in drinking water production and delivery or elsewhere, require significant care and handling. Required care and handling of the data are dictated by the sensor precision and accuracy, the intended use of the sensor (event detection, process control, research, compliance monitoring or other purposes), and the sensor context (their placement in the water production system, the variability in a monitored parameter at the sensor location, the difference between average conditions at a particular sensor location and a regulatory level or level of concern). This study presents an analysis of real-time water quality monitoring data management and analysis for sensors deployed in Philadelphia Water’s distribution system and used for event detection. First, the data collection, management and analysis tools are described and difficulties encountered and overcome while developing the system are highlighted. Those difficulties include both the complexity of collecting large volumes of data from a dispersed network of sensors and steps required to ensure data conform to requirements of event detection tools. Second, analyses are presented that quantify the required precision and accuracy of sensors deployed in the Philadelphia Water distribution system. Analyses include assessment in the variability of water quality data at several time scales and establishment of sensor performance metrics consistent with their use as part of an early warning system. In general, sensors in current use in Philadelphia Water’s sensor network were shown to meet required precision and accuracy for use as components in an event detection system. As shown in a companion paper to this paper, meeting the requirements was only possible after significant efforts to improve sensor installations and establish a rigorous operations and maintenance (O&M) effort. Finally, analyses for establishing sensor control limits are presented. Control limits are a critical input to event detection algorithms. To establish control limits, long time series of water quality data were analyzed and variability in the observations at the global, seasonal and daily time scales was characterized. Results indicate that control limits that optimize event detection performance (maximize detection of true events and minimize false positive and false negative detections) vary seasonally or possibly at shorter time scales.
Oral Presentation
Prepared by T. Bradley1, T. Bartrand2, R. Kopansky3, J. Rosen4
1 - Philadelphia Water - Bureau of Laboratory Services, need, need, need, need, United States
2 - Corona Environmental Consulting, 150 Monument Road, Bala Cynwyd, PA, 19004, United States
3 - Philadelphia Water, , , United States
4 - Corona Environmental Consulting, , , United States
Contact Information: [email protected]; 215-685-1460
ABSTRACT
Online sensor data, whether in drinking water production and delivery or elsewhere, require significant care and handling. Required care and handling of the data are dictated by the sensor precision and accuracy, the intended use of the sensor (event detection, process control, research, compliance monitoring or other purposes), and the sensor context (their placement in the water production system, the variability in a monitored parameter at the sensor location, the difference between average conditions at a particular sensor location and a regulatory level or level of concern). This study presents an analysis of real-time water quality monitoring data management and analysis for sensors deployed in Philadelphia Water’s distribution system and used for event detection. First, the data collection, management and analysis tools are described and difficulties encountered and overcome while developing the system are highlighted. Those difficulties include both the complexity of collecting large volumes of data from a dispersed network of sensors and steps required to ensure data conform to requirements of event detection tools. Second, analyses are presented that quantify the required precision and accuracy of sensors deployed in the Philadelphia Water distribution system. Analyses include assessment in the variability of water quality data at several time scales and establishment of sensor performance metrics consistent with their use as part of an early warning system. In general, sensors in current use in Philadelphia Water’s sensor network were shown to meet required precision and accuracy for use as components in an event detection system. As shown in a companion paper to this paper, meeting the requirements was only possible after significant efforts to improve sensor installations and establish a rigorous operations and maintenance (O&M) effort. Finally, analyses for establishing sensor control limits are presented. Control limits are a critical input to event detection algorithms. To establish control limits, long time series of water quality data were analyzed and variability in the observations at the global, seasonal and daily time scales was characterized. Results indicate that control limits that optimize event detection performance (maximize detection of true events and minimize false positive and false negative detections) vary seasonally or possibly at shorter time scales.
