An innovative approach to correct data from in-situ turbidity sensors for surface water monitoring

In-situ water quality sensors yield continuous, high-frequency measurements that are essential for long-term monitoring. Optical sensors, such as turbidity sensors, are prone to lens obstruction and ultimately, anomalous measurements. In this study, we developed a novel approach to detect, remove, and replace these anomalies. A free, open-source, software package implementing this approach is provided. Anomaly detection and removal is completely automated, though manual verification is recommended. Sections of missing data are filled by one of two methods, depending on the size of each section. This approach was tested on turbidity measurements from sensors deployed across southern Ontario. Automated detection performance was consistent across all test datasets, with most (>95%) synthetically introduced anomalies removed in all but one dataset. Gap-filling provided accurate estimates for smaller gaps, while performance on larger gaps varied. Overall, this approach as implemented in the provided software can greatly assist with data management in long-term monitoring programs.