Regional flood discharge modeling in Hamedan province using Artificial Neural Network

Flood is one of natural hazardous disasters that causes economic and life damages every year. Therefore, scientists have tried to assess the variability of this phenomenon. In this study, the ability of Artificial Neural Network (ANN) and Geographical Information System (GIS) in the estimation of flood discharges of 90 sub-basins of Hamedan Province (with concentration time less than 24 hours) was assessed using a data period of 16 years collected at the 17 hydrometric stations throughout the area. In this regard, area of the basins, elevation and mean slope, bindery of hydrologic groups of soil, weighted curve number as well as daily and 5 day precipitations which occured before the corresponding floods used as the input variables and the flood discharge defined as the output. Considering the training, validation and testing sets the results showed that the best structure was a feed forward ANN with two hidden layers of 5 and 4 processing elements. In this model, the coefficient of determination, root mean squared error and mean absolute error were 0.87, 0.72 and 2.83, respectively. Finally, using the maximum daily and 5 day precipitations for a return period of 25 years and application of the model, the spatial distribution of the direct runoff was predicted. Therefore, flood mapping of the province was specified to determine the priorities of flood control between the regions. Finally, the results indicated that ANN method was an appropriate tool for flood discharge modeling, especially in the case of missing data or inadequate hydrometric stations.