Optimization of cyanobacteria bloom management through improved forecasting models and optical sensors

Abstract

Cyanobacteria are a diverse group of microorganisms adapted to a range of environmental conditions that favour their ubiquity in waterbodies. Cyanobacteria harmful blooms (CyanoHABs) are events in which a cyanobacteria population grows rapidly, dominates the phytoplankton community and may release toxins or other noxious compounds in the waterbody. The frequency and magnitude of CyanoHABs are increasing as a result of climate change and increased pollution from urbanisation and agriculture expansion, representing a major risk to the public health and economy. Management of CyanoHABs is complicated by the temporal and spatial dynamic nature of these events, and by the large diversity of cyanobacteria species. Identification of the dominant cyanobacteria species is required to select appropriate mitigation and treatment alternatives. Therefore, water authorities have longed for reliable tools to support proactive and species-targeted CyanoHAB management. Emerging monitoring technologies and data-driven models represent a tangible opportunity to optimise CyanoHABs management by integrating rapid and taxa precise features into a single tool. Optical sensors, namely in-situ fluorescence sensors, allow rapid, remote estimation of the total phytoplankton and cyanobacteria concentration in a waterbody. This is done by measuring the fluorescence of the pigments chlorophyll a, common to all phytoplankton, and phycocyanin, exclusive to cyanobacteria. However, fluorescence estimates have limited taxa precision because they cannot discriminate between cyanobacteria species, and may have reduced accuracy, due to optical interferences. Data-driven models are increasingly being used to understand and predict complex ecological patterns, including cyanobacteria species succession, but the combination of high-frequency fluorescence data with data-driven models to optimise CyanoHAB management has seldom been investigated. The aim of this doctoral thesis is to develop an integrated model able to optimise CyanoHAB management by incorporating site-specific drivers of cyanobacteria succession and factors that affect fluorescence sensor estimates. This aim was achieved by addressing four objectives: (1) to systematically review the state-of-knowledge of forecasting and predictive CyanoHAB models and their application to freshwater lakes; (2) to test and quantify interferences, if any, on fluorescence probe measurements according to diel light variability and species composition; (3) to identify and quantify, through observational data analysis, dominance of cyanobacteria species according to site-specific environmental conditions; and (4) to establish a framework for implementation of integrated models considering fluorescence sensor calibration and prediction of cyanobacteria species succession. This research project’s combination of observational data analysis and analytical laboratory work underpins its novelty and relevance. Observational data analysis was performed for three Australian drinking-water reservoirs and correlations between environmental drivers and dominance of key cyanobacteria species were determined for Wivenhoe Lake (Queensland), Tingalpa Reservoir (Queensland) and Myponga Reservoir (South Australia). Two sets of controlled laboratory experiments were then performed. The first experiment analysed the fluorescence characteristics of four key cyanobacteria species (Aphanocapsa sp., Microcystis aeruginosa, Dolichospermum circinale and Raphidiopsis raciborskii) that are often dominant in the assessed drinking-water reservoirs. The experiment quantified the variability of the species’ fluorescence characteristics throughout their respective growth phases and also compared the differences among morphologically similar species. The second experiment analysed light-induced quenching in a cyanobacterium (Dolichospermum variabilis) and a green alga (Ankistrodesmus gracilis) by simulating diel light variability under controlled temperature and stratification conditions. Lastly, a framework combining the methodological procedures from the observational data analysis and the fluorescence calibration experiments was established with the aim of supporting the development of species-targeted models utilizing fluorescence sensors. An integrated model based on the framework was developed and tested in Myponga Reservoir, South Australia. Moreover, a continuous improvement process for CyanoHAB models and guidelines of best practices for fluorescence sensors deployment, calibration and operation were developed as a result of this research. The methods and findings are provided in four peer-reviewed journal papers included as chapters in this thesis (i.e., chapters 3, 5, 6 and 7) and a final discussion chapter (chapter 8). Objective 1 findings revealed that high-frequency data, such as the data from optical sensors, can improve performance of CyanoHAB models. For Objective 2, two key findings should be highlighted. First, fluorescence per cell was found to significantly vary among species, while fluorescence per unit of biomass (estimated from biovolume) was much more consistent among species. Second, diel light variability reduced fluorescence for both cyanobacteria and green algae up to 79% under the assessed conditions. Objective 3 findings indicated that environmental drivers for cyanobacteria succession and dominance are mostly site-specific. Species-specific traits, such as diazotrophy and gas vesicles, interact in complex ways with local environmental conditions leading to variable dominance succession among species. Finally, the key findings of Objective 4 showed that the required steps to develop a species-targeted CyanoHAB model using fluorescence sensors are feasible, given that constraints in data availability are met. Overall, the findings of this PhD research indicate that CyanoHAB management can be optimised through the combination of fluorescence sensors and forecasting models based on data-driven approaches, as long as rigorous calibration and data analysis procedures are undertaken. Importantly, the findings also highlight that even though generalisable patterns of species-specific drivers exist, site-specific analysis is required due to the complex interactions between the several factors involved in the occurrence of CyanoHABs.