Modeling and forecasting of runoff using fuzzy logic (ANFIS) and physical based (HEC-HMS) models

In this study, Kurukavak stream basin was selected as study basin, and flow modeling and forecasting are performed by using data of storm events that took place on 30.06.1988 and 01.06.2005. ArcGIS (Version 10.3.1) software was used to characterize the basin. Specified characteristics are classified and transferred to the maps. Two different models are included in the modeling studies; namely; HEC - HMS model (HEC - HMS Version 4.2.1 software) and ANFIS model (MATLAB Version R2018a) and in both calibration and validation are performed. The results of HEC - HMS and ANFIS models and the performance of harmony between these two models are evaluated in the conclusions and recommendations section. According to the results, the calculated NSE values in the HEC - HMS model were 0.824 and 0.618 for calibration and validation periods respectively. The evaluation based on the NSE method states that, the event data used during the calibration - validation periods are at certain intervals and demonstrate the reliability of the modeling. Additionally, in the test phase of the ANFIS model, the mean square error (RMSE), was calculated as 0.019542. It has been observed that, HEC - HMS and ANFIS models have good performances and predicted runoff values are close to observed values. The good and similar results of these two models have shown that the ANFIS model can be used reliably in cases where the measurement of basin characteristic parameters cannot be performed or is missing.