Prediction of lake eutrophication using artificial neural networks

An artificial neural network (ANN), which is a data-driven modelling approach, is proposed to indicate the water quality of Lake Fuxian, the deepest lake of southwest China. To determine the nonlinear relationships between the water quality factors and eutrophication indicators, several ANN models were chosen. The back-propagation and radial basis function neural network models were applied to relate the key factors that influence a number of water quality indicators, such as total nitrogen (TN), secchi disk depth (SD), dissolved oxygen (DO) and chlorophyll-a (Chl-a) in Lake Fuxian. The measured data were fed to the input layer, representing forcing functions to control the in-lake biochemical processes. Eutrophication indicators (TN, SD, DO and Chl-a) were represented in the output layers. The results indicated that the back-propagation neural network model performed better than radial basis function neural network model in ten months prediction and was able to predict these indicators with reasonable accuracy. Such neural networks can be a valuable tool for lake water management.