Developing cyanobacterial bloom predictive models using influential factor discrimination approach for eutrophic shallow lakes

Abstract

Harmful cyanobacterial blooms damage aquatic ecosystems and pose a threat to human health. To identify key factors causing cyanobacterial blooms in eutrophic shallow lakes, we analyzed cyanobacterial and physicochemical water samples of 12 sites collected monthly from December 2012 to December 2019 in Dianshan Lake. We found that the rapid growth of cyanobacteria was limited by a temperature threshold. When the air temperature was below 18 °C, the sampled physicochemical factors could not make difference in cyanobacterial abundance regardless the values of these parameters. However, when the air temperature was above 18 °C, the measured physicochemical factors played important roles in influencing cyanobacterial abundance. We developed a data-driven predictive model for cyanobacterial blooms based on seven-year data from Dianshan Lake using multiple logistic regression. Such a model could be easily used to predict cyanobacterial blooms. Our weight analysis of model parameters indicated that dissolved substances other than TN and TP are the key factor determining cyanobacterial blooms in nitrogen and phosphorus rich shallow freshwater lakes once air temperature is above 18 °C. Eutrophic shallow lakes are prone to cyanobacterial blooms, and unwashed data analysis may mask key factors determining cyanobacterial blooms, which obscures the prediction of cyanobacteria blooms. Our results are helpful to uncover the real causes of the blooms of eutrophic shallow lakes in China and elsewhere, and hence improve the understanding and management in controlling cyanobacterial blooms.