HYDRONEURALNETWORKS: A SOFTWARE FOR HYDROLOGICAL MODELING USING ARTIFICIAL NEURAL NETWORKS

Artificial neural networks (ANNs) are a new tool for hydrological sciences where research is underway to develop and discover new applications that can be applied in hydrology. ANNs are capable of modeling data with functional relationships that are not known in advance. They have been recognized as very powerful tools in capturing simple as well as complex functional relationships between several input and output variables. Many studies have shown the potential of ANNs for modeling rainfall-runoff relationships over watersheds and comprise most ANN applications today. Unfortunately there is a lack of computer applications for hydrological modeling using ANNs with a user-friendly interface. In this study, we developed HydroNeuralNetworks, a user-friendly software tool developed for hydrological modeling that runs under Matlab 2011a and above. The software uses a multilayer feedforward network. This architecture can approximate any nonlinear function. Feedforward networks consist of a series of layers where the first layer has a connection from the network input. Each subsequent layer has a connection from the previous layer. The final layer produces the network's output. The software also presents many choices to customize the network and includes 12 training algorithms and many statistical error estimators and performance function to improve the quality of the simulation. To give an overview of the scope of the software a case study for daily streamflow forecasting using ANNs is presented under HydroNeuralNetworks.