Online forecasting chlorophyll a concentrations by an auto-regressive integrated moving average model: Feasibilities and potentials

Algal blooms are commonly observed in freshwater and coastal areas, causing significant damage to drinking water and aquaculture production. Predictive models are effective for algal bloom forecasting and management. In this paper, an auto-regressive integrated moving average (ARIMA) model was developed to predict daily chlorophyll a (Chl a) concentrations, using data from Taihu Lake in China. For comparison, a multivariate linear regression (MVLR) model was also established to predict daily Chl a concentrations using the same data. Results showed that the ARIMA model generally performed better than the MVLR model with respect to the absolute error of peak value, root mean square error and index of agreement. Because the ARIMA model needs only one input variable, it shows greater applicability as an algal bloom early warning system using online sensors of Chl a. (C) 2015 Elsevier B.V. All rights reserved.