Intelligent system for remotely monitoring manganese concentrations in water reservoirs: A case study of a water quality-monitoring program at Advancetown Lake in South-East Queensland

Abstract

ontinuously monitoring and managing manganese (Mn) concentrations in drinking water reservoirs is of paramount importance for water suppliers, as high soluble Mn levels can lead to the discoloration of potable water. Traditional Mn management involves regular manual water sampling and laboratory analyses. In cases where critical Mn concentration thresholds are exceeded, appropriate treatment procedures are adopted. Despite the Mn level currently being manually sampled throughout the year, in many subtropical monomictic lakes - such as Advancetown Lake on the Gold Coast - Mn concentrations in the epilimnion, where the water is drawn for potable use, are usually only elevated during winter, with the onset of partial or full lake destratification. Vertical profiling systems (VPS) have been installed in Seqwater's stored water reservoirs to continuously collect physical parameters such as: water temperature; specific conductivity; turbidity; pH; REDOX; chlorophyll-a, blue-green algae; and dissolved oxygen. These may be used to accurately determine the transport processes of Mn within the lake system. Therefore, a historical database of VPS and Mn laboratory testing data provides the opportunity to develop a data-driven prediction model that can autonomously forecast seven days in advance the Mn concentrations at the drawn-off depth for water treatment plants. In this study, a VPS was employed alongside physically collected water quality data, and analysed to deliver data-driven predictive models associated with the real-time VPS data collection. These models were able to forecast future Mn concentrations up to seven days ahead with correlation coefficients higher than 0.83 for an independent test dataset. Importantly, the peak concentrations in the epilimnion during lake destratification were predicted with correlation coefficients of greater than 0.90. The models also display the probabilities of the Mn to exceed critical thresholds, thus assisting operators in Mn treatment decision-making. Such a tool is highly beneficial for water suppliers, as the cost and time spent monitoring Mn concentrations can be significantly reduced and more proactive forecasting and planning for elevated levels of Mn can be enabled.