Assessment of the Capability of Hydrologic and Artificial Neural Network Models for Flood Warning System in Land Use Change Condition

The main purpose of this research is comparing the capability of HEC-HMS model and dynamic artificial neural networks (DANN), forecasting lead time of Golabdare-Darband basin in the north of capital city Tehran. Five scenarios are defined for different land uses, including conditions of 1955, 1988 and 2001, suitable and unsuitable land use of the basin. Floods with different return periods are simulated. The result of HEC-HMS model is used as training and test data of ANN network. The result of artificial neural network showed that the number of rainfall and runoff delays has optimal. Comparison of forecast lead time in two models shows that the lead time increases by decreasing in returned periods. Comparison of forecast lead time of five scenarios shows the scenario of unsuitable management of the watershed has the shortest lead time by using both models. The longest forecast lead time is for the scenario of 1988 land use by using DANN model and the scenario of 1955 land use for HEC-HMS model. The difference comes from instability of simulation by DANN.