GIS Based Artificial Neural Network(Ann) Method for Flood Susceptibility Mapping Case of Djelfa City (Algeria)

Abstract

Abstract Djelfa city situed in the center of Algeria,is particuulary prone to the risk of flooding due to its topography and location,especially considering the presence of unpredictable and forceful watercourses like Oued mellah and Oued boutrifis flowing through the urban area. Various methods exist for predicting and mapping flood susceptibilityand the latest approaches involve deep learning machine and artificial neural networks,which were employed in the current study. Four geoenvironmentalflood conditioning factors were considered including elevation, slope,urban density and distance to streams. Recent artificial neural network(ANN)model has been used to obtain an optimal output with minimized cross entropy error and better assess flood susceptibility in Djelfa City.The weights for each factor were determined using the backpropagation training method.Subsequently,flood susceptibility indices were calculated using the trained backpropagation weights and susceptibility maps were created based on geographic information system (GIS) data.The results of the flood susceptibility maps were then compared to flood location data to validate the model.The good convergence of the resultsclearly demonstrates that artificial neural network is an effective tool to analyzing flood susceptibility.