River Flow Estimation Using Feed Forward and Radial Basis Neural Networks Approaches

In this study, applicability of Artificial Neural Network (ANN) methods for river flow estimation is investigated. Accurate modeling and forecasting of hydrological processes such as rainfall, rainfall-run off relationship, runoff, is important for management and planning of water resources. To illustrate the capability of ANN for modeling of water resources, Great Menderes River, locate in the west of Turkey, is chosen as case study area and daily river flow estimation study is carried out. For this aim, totally 4383 daily river flow data are obtained from river flow gauges station of Great Menderes Çıtak Bridge (713) at period 1988 – 2000 years and the models having various input structures are constructed. The two different types of ANN, Feed Forward Neural Networks and Radial Basis Neural Networks, are used to estimate daily river flow. The models are trained and tested by FFNN and RBNN and the results of models are compared with field observation data. The criteria of performance evaluation are calculated in order to evaluate and compare the performances of FFNN and RBNN models. Then the best fit model and network structure are determined according to these criteria.