Rainfall-Runoff Modeling by Adaptive Neuro-Fuzzy Inference System (ANFIS) and Multi-Variable Linear Regression (MLR)

In this research, the efficiency of Neuro-Fuzzy System was evaluated for estimating runoff in the mountainous area of the Haraz watershed. The goal is to create a model with proper functions and membership degrees that can properly estimate the relationship between rainfall and runoff in either basin or watershed. Thus, to predict the amount of runoff, 44 different combinations of rainfall, temperature, evaporation, flow rate and antecedent precipitation index with lag time entered to an ANFIS model during the period of 32 years as a daily data. Among the different combinations of input, the structure of rainfall and mean temperature of current day with moisture index and flow rate of 1-day ago was selected as the appropriate model. The results of t-test for determining the significant difference showed that there is no significant difference between ANFIS and MLR methods. The ANFIS system with a two degree triangular membership function with RMSE=2.67, VE=4.16, R2=0.98 and EF=0.97 compared to the MLR with RMSE=2.83, VE=4.55, R2=0.94 and EF=0.92 during the test period, has a better performance in rainfall-runoff modeling of the mountainous area of the Haraz watershed.