Evaluation of two combined hydrological-black box models for flood forecasting in Halilrud river basin

Forecasting river flow in flood conditions is an essential part of river engineering studies that, if accurately estimated, can greatly contribute to the effectiveness of management programs and reduce financial and human losses. Using appropriate models and increasing the accuracy of these models will lead to improvements in the accuracy of prediction results. One of the available solutions to increase the accuracy of existing and valid rainfall-runoff models is to build a combined model with the help of preferably intelligent methods link to these models. In the present study, the combination of WEAP 21 hydrological model with black box models based on ANN and GMDH methods is used to increase the accuracy of WEAP model and then the created model was used to simulate the flood of part of Halilrud river in Kerman province. Precipitation, flow, humidity, wind and temperature data were entered into the WEAP model for the existing record minus the last two years. After calibrating and validating the model, the last two years were forecasted. The results showed that the WEAP-ANN (R2 = 0.78) model was able to estimate the runoff with higher accuracy compared to the WEAP-GMDH (R2 = 0.59) and WEAP (R2 = 0.14) models in the basin.