This article has been peer reviewed and accepted for publication. It is in production and has not been edited, so may differ from the final published form.
Prediction of cyanobacterial blooms in the Dau Tieng reservoir using artificial neural network
An artificial neural network (ANN) model was applied to predict the cyanobacteria bloom in the Dau Tieng Reservoir, Vietnam. Eight environmental parameters (pH, Dissolved oxygen-DO, Temperature, Total dissolved solids-TDS, Total Nitrogen-TN, Total Phosphorus-TP, Biochemical oxygen demand-BOD5, and Chemical oxygen demand-COD) were introduced as the inputs, whereas cell density of 03 cyanobacterial genera (Anabaena, Microcystis, and Oscillatoria) with microcystin concentrations (MCs) were introduced as the outputs of the three-layer feed-forward back-propagation ANN. Eighty networks covering all combinations of 04 learning algorithms (Bayesian regularization-BR, Gradient descent with momentum and adaptive learning rate-GDX, Levenberg Mardquart-LM, Scaled conjugate gradient-SCG) with 02 transfer functions (tansig, logsig) and 10 numbers of hidden neurons (6-16) were trained and validated to find the best configuration fitting the observed data. The result is a network using BR learning algorithm, tansig transfer function and 9 neurons in the hidden layer showing satisfactory predictions with the low values of error (RMSE=0.108) and high values of correlation coefficient (R=0.904) between experimental and predicted values. Sensitivity analysis on the developed ANN indicated that TN and Temperature could have the most positive and negative effect, respectively, on MC value. These impressive results indicate that ANN modeling could effectively predict the behavior of the cyanobacteria bloom process.
MF16327 Accepted 12 March 2017
© CSIRO 2017