Fault Detection for Salinity Sensors in the Columbia Estuary

Sensors developed in the Columbia River estuary gather information on physical dynamics and changes in estuary habitat. Of these sensors, conductivity sensors are particularly susceptible to bio-fouling, which gradually degrades sensor response and corrupts critical data. Several weeks may pass before degradation is visibly detected. As a result, an indeterminate amount of the archival data is corrupted, as the onset time of bio-fouling is unknown. To speed detection and minimize data loss, we develop automatic bio-fouling detectors, based in machine learning approaches, for these conductivity sensors. We demonstrate that our detectors identify bio-fouling at least as relably as human experts and provide accurate estimates of bio-fouling onset time. Real-time detectors installed during the summer of 2001 produced no false alarms, yet detected all episodes of sensor degradation before the field staff.