Application of Artificial Neural Network in the Research of the Bohai Bay Eutrophication

In order to research the feasibility of artificial neural network (ANN) in the research of eutrophication of the Bohai Bay in China, an ANN model simulating chlorophyll a, b and c concentrations, concerning temperature, dissolved oxygen, salinity, pH value, chemical oxygen demand (COD), PO43-, NO2-and NO3-factors in the Bohai Bay was presented and validated. After experiencing and training by Matlab, the model′s validation mean square error (MSE) performance is 0.009 985 02. R-squared between estimated and observed concentrations of chlorophyll a, b and c are 0.965 7, 0.998 7 and 0.970 7 respectively, indicating that the estimated value agrees with the observed value well, and the model can be used in the prediction of eutrophication of the Bohai Sea. In order to study the influence of model input factors on chlorophyll concentration (i.e. model outputs), hypothetical scenarios were introduced to show model output responses to variations in input factors. The limitation of temperature, salinity and phosphate that induce red tide in the Bohai Bay was also presented.