Predicting cyanobacteria occurrence using climatological and environmental controls.

The occurrence of algal bloom results in deterioration of water quality, undesirable sights, tastes and odors, and the possibility of infections to humans and fatalities to livestock, wildlife and pets. Earlier studies have identified a range of factors including water temperature, flow, and nutrient concentrations that could affect cyanobacterial proliferation. Lack of enough data, independence in data across multiple sampling time steps, as well as the presence of more than one causative factors, each with different levels of influence on the response, has resulted in limited progress in the development of generalized prediction frameworks for cyanobacteria. In this study, a prediction model for cyanobacteria occurrences was developed using only three dominant environmental variables; water temperature, velocity and phosphorus concentration. These environmental variables were selected due to not only direct or joint contribution to algal bloom but also the ease of their availability either through direct measurements or as modelled responses in the river location of interest. In order to apply bacterial growth dynamic to the model, weight functions which quantify the importance assigned to the three variables depending on the cell number at the preceding time, were formulated. An extensive dataset spanning from 2013 to 2018 at 16 representative locations across the four major rivers in South Korea was used to develop and validate the model. Through cross-validation, this model was shown to have more than 75% forecasting accuracy despite the use of a relatively simple predictive algorithm. As the developed model makes use of commonly available environmental variables, it can easily be extended to locations across the country where very limited or no prior information about cyanobacteria bloom is available.