Rainfall–runoff modeling using adaptive neuro-fuzzy inference system (ANFIS) and genetic algorithm (GA)

Abstract Nonlinear properties and natural uncertainties in the rainfall–runoff process, necessity of extensive data, complexity physical models have caused researchers to use methods inspired by nature such as artificial neural networks, fuzzy systems, genetic algorithms (GA). The main purpose this study was estimate runoff employing Adaptive Neuro-Fuzzy Inference System (ANFIS) GA using accessible, applicable, easily available climatic data. results two were compared provide an easy but reliable model evaporation. utilized Sivand river basin located Fars province central Iran. considering a range performance indicators mean absolute error (MAE), Nash–Sutcliffe efficiency coefficient (NSE), root square (RMSE), correlation (R2). According presented, ANFIS with lower RMSE MAE higher NSE between observed predicted values provided accuracy comparison GA. Also, it clear that ANFIS, increase number membership functions running cycles decreased studied stations improved 42, 44 11%, respectively increasing run rounds. nonlinear performed better than linear when applying non-linearizing three 27.5%, 17%, 9.5%, respectively. Meanwhile, amounts best from 23%, 54.6%, 35.7% stations, can be estimated appropriately utilizing ony meteorological data there is no need for more complex interdependent A sensitivity analysis conducted too removing rainfall evaporation parameters different scenarios. showed lowest absence those especially scenario 3 = 0, ?0.005 Chambian, Dashtbal, Tang Balaghi justifies only areas scant are available.