APPLICATION OF FUZZY NEURAL NETWORK MODELING TECHNIQUES FOR ESTIMATION OF RUNOFF IN BLUE NILE RIVER BASIN A CASE STUDY OF KOGA CATCHMENT

Abstract

Runoff estimation from a catchment is a vital phase in water resource planning and management. Rainfall plays the major role in computing runoff from a catchment along with other catchment processes. In the past few decades, a wide variety of automated or Computer-based approaches have been applied to model this rainfall­ runoff process. However, many such approaches have an important limitation in that they treat the rainfall-runoff process as a realization of only a few parameters of linear relationships rather than the process as a whole. What is required, therefore, is an approach that can capture not only the overall appearance but also the intricate details of the nonlinear behavior of the process. In this study, the applicability of fuzzy neural network modeling techniques for proper estimation of runoff was investigated. The proposed Fuzzy Neural Network is a hybrid combination of Fuzzy Logic and Artificial Neural Network, Which are complimenting each other. Here, effective rainfall and runoff of various lag periods were used as input Fuzzy variables to organize knowledge that is expressed 'linguistically' into a formal analysis. By applying fuzzy neural network, Fuzzy rule base was formulated by classifying input values into various fuzzy sets. The neural network technique has been used to train the sets which were equipped with fuzzy information in the proposed FNN model. For this purpose, after identifying the universe of discourse, the whole data range has been fuzzified into small subintervals with suitable overlaps. ANN modeling was done with subintervals (fuzzy set) for driving membership functions of fuzzy sets incorporating rule base knowledge. Finally defuzzification was done to obtain crisp results in terms of runoff. Based on the modeling results for runoff estimation in koga catchment, it is concluded that Fuzzy Neural Network has a promising potential for providing reliable runoff estimation than single Artificial Neural Network Models.