Zoning avalanche-prone sirvan basin with the combination of hierarchical analysis and artificial neural networks

One of the concerns of people in the mountainous areas is snow avalanche. In this article avalanche zoning as a hazard using artificial neural networks and hierarchical analysis has been acting. AHP to improve the training sample is conducted in GIS. This method applied at basin sirvan in the South West of kurdistan province that have high potential for avalanche risk,. For this purpose, First it was necessary avalanche pathways that in them avalanche was happened, Field visits were conducted and their coordinates were taken. A literature review was conducted to identify factors affecting this process. Based on studies slope, aspect, elevation, convexity and concavity, distance from roads and land use were selected. Map of hierarchical classification analysis of each class of 20 samples were used to train the neural network. Perceptron neural network to assess these variables with the six input layer, a hidden layer, six nodes per layer with learning rate 0.01 with two linear sigmoid function as the optimal structure by trial and error be accepted. Evaluation of these variables using neural networks shows that more than 86 percent of the study area is among the areas with high potential risk of avalanche. In order to validate these models from observational data available that demonstrates the success and effectiveness of both function but with low priority for linear function.