Optical Sensors and Machine Learning for Optimised Cyanobacteria Bloom Management

Cyanobacteria blooms can be a threat to human health. Drastic changes in cyanobacteria concentrations are often rapid; thus high-frequency sampling is required for effective bloom detection and control. Traditional monitoring methods are unsuited to such analysis and can be onerous. On the other hand, in-situ fluorescence probes are currently employed by a number of water utilities in order to obtain near real-time water quality monitoring. Nevertheless, to date fluorescence probes can only derive rough estimates of total cyanobacteria concentrations, without discretizing bloom composition to species/genus level. The knowledge of the bloom species’ composition can be helpful since some species are known toxin and/or taste and odour producers. The literature shows that different species of cyanobacteria have different tolerances for environmental conditions such as water temperature, pH and nutrition status. In this research, we argue that approaches such as machine learning can be applied to explain and predict species-specific cyanobacteria blooms based on environmental conditions. In particular, the feasibility of a model to predict cyanobacteria bloom species composition through a combination of real-time fluorescence biological data and routine environmental conditions data is discussed. As an example, preliminary analysis from an Australian drinking water reservoir was conducted and showed patterns of dominance and appearance of key species according to environmental conditions. Further, laboratory work combined with fluorescence monitoring of the reservoir indicates potential to couple data-driven species predictions with real-time total cyanobacteria monitoring through calibration. The development of such a model could be useful for decision-makers to evaluate risk and establish proper management responses.