Spatial predication of flood zonation mapping in Kan River Basin, Iran, using artificial neural network algorithm

Abstract

Flood is one of the significant natural disasters, which is treated as one of the main global concerns which increased occurrence has led to an increase in mortality rates and economic losses. Various methods have been developed and proposed for the analysis of this natural disaster. Iran is among several countries in the world, which faces severe problems of flood each year particularly in urban catchments. The present study aims to utilize GIS spatial analysis functions, data from Hydrometric and Rain-Gauge stations, satellite images, and thematic data layers in the form of Artificial Neural Network Algorithm for prediction of discharge values and spatial modeling of floods in Kan River Basin located in Tehran province. An optimized artificial neural network of 7 inputs, including slope, slope curvature, flow accumulation, NDVI, geological units, soil type, and rainfall data along with eight, sixteen and one neurons for the first, second and output hidden layers, respectively, were designed and developed. The output of the neural network was discharge values in stations. According to Table .2 in the result section, ANN method has one of the highest correlation and lowest RMSE in flood modeling. Precision parameters such as R2, RMSE and MAE were used to show the efficiency of the proposed model which yielded the values of 0.82, 0.18, and 0.13, respectively. The results obtained by the present study can be employed in future environmental planning at local scale as a means for improving the management of environmental risks and crises. The present study showed that an integrated utilization of GIS spatial analysis function with neural network algorithm is one of the high efficiency methods for predicting the potential of natural disasters such as floods.