Application of Several Data-Driven Techniques for Rainfall-Runoff Modeling

In this study, several data-driven techniques including system identification, adaptive neuro-fuzzy inference system (ANFIS), artificial neural network (ANN) and wavelet-artificial neural network (Wavelet-ANN) models were applied to model rainfall-runoff (RR) relationship. For this purpose, the daily stream flow time series of hydrometric station of Hajighoshan on Gorgan River and the daily rainfall time series belonging to five meteorological stations (Houtan, Maravehtapeh, Tamar, Cheshmehkhan and Tangrah climatologic stations) were used for period of 1983-2007. Root mean square error (RMSE) and correlation coefficient (r) statistics were employed to evaluate the performance of the ANN, ANFIS, ARX and ARMAX models for rainfall-runoff modeling. The results showed that ANFIS models outperformed the system identification, ANN and Wavelet-ANN models. ANFIS model in which preprocessed data using fuzzy interface system was used as input for ANN which could cope with non-linear nature of time series and performed better than others.