Prediction of Harmful Algal Bloom Using Machine Learning in Paldang Lake, South Korea

The risk of algal bloom is increasing due to climate change and an increase in the inflow of pollutants into the water system by agricultural practice and industrial development. Paldang Lake in the Han River passing through the metropolitan area of South Korea plays an imperative role as a major water supply for citizens in many big cities. Although Harmful Algal Bloom (HAB) prediction is essential for managing water supply sources in this metropolitan area, however, there are very few studies on Paldang Lake. Therefore, in this study, we construct a machine learning-based algal bloom prediction model of Paldang Lake using several different machine learning models including the Support Vector Machine (SVM) algorithm and Long Short-Term Memory (LSTM). This prediction model of HAB is constructed using a set of big data including historical meteorological observations, water quality data, and hydrological big data as input data, and it produces HAB cell count data as output data. In addition, we apply future climate projection to the constructed model to examine future climate change impacts on future HAB events. It is expected that the prediction model of algal bloom in Paldang Lake, developed from this study, will be available as an indicator of establishing criteria for non-point pollution sources and water quality management plans near Paldang Lake. Furthermore, this model can be used as a testbed to develop and test physically-based various surface hydrological and water quality models.