Early warning of algal blooms based on the optimization support vector machine regression in a typical tributary bay of the Three Gorges Reservoir, China

Algal blooms caused by climate change and human activities have received considerable attention in recent years. Since chlorophyll a (Chl-a) can be used as an indicator of phytoplankton biomass, it has been selected as a direct indicator for monitoring and early warning of algal blooms. With the development of artificial intelligence, data-driven approaches with small sample data and high accuracy prediction have been gradually applied to water quality prediction. This study aimed at using environment factors (water quality and meteorological data) to assist the prediction of Chl-a concentration based on the optimization support vector machine (SVM) model. The most relevant environment factors were extracted from the commonly used environment factors according to the method of cosine similarity. The traditional particle swarm optimization (PSO) algorithm was adopted to optimize the ANN and SVM models, respectively. Then, the better prediction model PSO-SVM can be obtained according to the results of three scientific evaluation indicators. The latest optimization algorithm of grey wolf optimizer (GWO) was also proposed to optimize the SVM to realize high-accuracy Chl-a concentration predication. The GWO-SVM model achieved higher accuracy than the other models both in training and validation processes. Therefore, the dimension of the input vector could be reduced with using the cosine similarity method, and the prediction of Chl-a concentration in high accuracy and the early warning of algal blooms in the study area of this paper could also achieved.