Earthquake risk assessment using an integrated Fuzzy Analytic Hierarchy Process with Artificial Neural Networks based on GIS: A case study of Sanandaj in Iran

Abstract Earthquakes are natural phenomena, which induce natural hazard that seriously threatens urban areas, despite significant advances in retrofitting urban buildings and enhancing the knowledge and ability of experts in natural disaster control. Iran is one of the most seismically active countries in the world. The purpose of this study was to evaluate and analyze the extent of earthquake vulnerability in relation to demographic, environmental, and physical criteria. An earthquake risk assessment (ERA) map was created by using a Fuzzy-Analytic Hierarchy Process coupled with an Artificial Neural Networks (FAHP-ANN) model generating five vulnerability classes. Combining the application of a FAHP-ANN with a geographic information system (GIS) enabled to assign weights to the layers of the earthquake vulnerability criteria. The model was applied to Sanandaj City in Iran, located in the seismically active Sanandaj-Sirjan zone which is frequently affected by devastating earthquakes. The Multilayer Perceptron (MLP) model was implemented in the IDRISI software and 250 points were validated for grades 0 and 1. The validation process revealed that the proposed model can produce an earthquake probability map with an accuracy of 95%. A comparison of the results attained by using a FAHP, AHP and MLP model shows that the hybrid FAHP-ANN model proved flexible and reliable when generating the ERA map. The FAHP-ANN model accurately identified the highest earthquake vulnerability in densely populated areas with dilapidated building infrastructure. The findings of this study are useful for decision makers with a scientific basis to develop earthquake risk management strategies.