A Neural Data Fusion Model for Hydrological Forecasting

A model of adaptive data fusion flow forecasting using concurrent neural networks (ADAFCON) is presented. It uses a fusion of rainfall and discharge data . The system consists of three backpropagation neural networks. Each neural module is trained to estimate a specific class of data dynamics: low, medium and high gradients. The decision fusion module uses a concurrential strategy. The model is applied to forecast the Piura river flow (discharge).