عنوان مقاله [English]
نویسندگان [English]چکیده [English]
The occurrence of horrific floods as a result of climatic changes in recent decades has caused many damages in different areas of the world. In dry areas, the effects of these changes are more tangible. Among these regions, Sistan and Baluchestan province, regarding its warm and dry climate, is flood prone. Sarbaz basin which is located in the southern parts of this vast province, each year is encountered with flood occurrences and its destructive consequences under the affection of the current situation. This aim of this study is to forecast the floods of Sarbaz River via an artificial neural network.
In this study, three networks including multi-layer perception network, back propagation and radial basis were used to predict floods of Sarbaz River and the results of these networks were compared through multiple regression models. To this aim, the data of three daily climatic and hydrological stations of Sarbaz, Pirdan and Iranshahr were used over a period of 28 years (October 1981 upto Septenber 2009). Analyzing the correlation between the data and the flow rate of Sarbaz River, the effective parameters on floods were determined. After normalizing the data, different models were created. Surveying the results showed that the selected network Radial Basis with the correlation 0.97 at the training phase and 0.714 at testing phase and less error comparing with other networks, was selected as the best model among various neural networks. Comparing these network results and regression models showed that the neural network model has a better performance and presents a better prediction of Sarbaz river flood using the regression method.