Near real-time event detection for watershed monitoring with CANARY

Abstract

Illicit discharges in surface waters are a major concern in urban environments and can impact ecosystem and human health by introducing pollutants (e.g., petroleum-based chemicals, metals, nutrients) into natural water bodies. Early detection of pollutants, especially those with regulatory limits, could aid in timely management of sources or other responses. Various monitoring techniques (e.g., sensor-based, automated sampling) could help alert decision makers about illicit discharges. In this study, a multi-parameter sensor-driven environmental monitoring effort to detect or identify suspected illicit spills or dumping events in an urban watershed was supported with a real-time event detection software, CANARY. CANARY was selected because it is able to automatically analyze data and detect events from a range of sensors and sensor types. The objective of the monitoring project was to detect illicit events in baseline flow. CANARY was compared to a manual illicit event identification method, where CANARY found > 90% of the manually identified illicit events but also found additional unidentified events that matched manual event identification criteria. Rainfall events were automatically filtered out to reduce false alarms. Further, CANARY results were used to trigger an automatic sampler for more thorough analyses. CANARY was found to reduce the burden of manually monitoring these watersheds and offer near real-time event detection data that could support automated sampling, making it a valuable component of the monitoring effort.