Prediction of Runoff Coefficient under Effect of Climate Change Using Adaptive Neuro Inference System

Abstract

The complex characteristics of the rainfall- runoff mechanism, along with its non-linear attributes and inherent uncertainties, have prompted scholars to explore alternative approaches inspired by natural phenomena. In order to tackle these obstacles, artificial neural networks (ANN) and fuzzy systems (FL) have been utilised as feasible substitutes for conventional physical models. Furthermore, the procurement of comprehensive data is considered essential for precise analysis and modelling. This study's primary objective was to use pertinent climatic data such as; Precipitation (P), Temperature (T), Relative humidity (Rh), and Wind speed (Ws) to predict the runoff coefficient using the Adaptive Neuro-Fuzzy Inference System (ANFIS) model. Different ranges (60:40; 70:30; 80:20) were used for the training and testing phases. The model was employed to predict the runoff coefficient in the Aksu river basin in Antalya province in Turkey. The study conducted a comparative analysis of the results, taking into account various performance indicators of the model, such as mean absolute error (MAE), Nash-Sutcliffe efficiency coefficient (NSE), root mean square error (RMSE), and correlation coefficient (R²). Based on the findings presented, the (60:40) range showed the best results as evidenced by its low RMSE and MAE values and its high R2 and NSE values (RMSE:0.056, MAE:1.92, NSE:0.868, R2 :0.996). It was concluded that the ANFIS model magnificently predicts runoff coefficients with an exceptional level of precision, also the study findings indicate that accurate runoff coefficient estimation can be achieved using meteorological data without incorporating more intricate and interrelated data.