An optimization of artificial neural network model for predicting chlorophyll dynamics

As one of the factors to represent some species of algae, chlorophyll dynamics model has been regarded as one of the early-warning proactive approaches to prevent or mitigate the occurrence of some algal blooms. To decrease the cost of aquatic environmental in-situ monitoring and increase the accuracy of bloom forecasting, a traditional artificial neural network (ANN) based chlorophyll dynamics prediction model had been optimized. This optimization approach was conducted by presenting the change of chlorophyll value rather than the base value of chlorophyll as the output variable of the network. Both of the optimized and traditional networks had been applied to a case study. The results of model performance indices show that the optimized network predicts better than the traditional network. Furthermore, the non-stationary time series was employed to explain this phenomenon from a theoretical aspect. The proposed approach for chlorophyll dynamics ANN model optimization could assist the essential proactive strategy for algal bloom control.