Evaluation of ANFIS and LR models for seismic rockfalls’ susceptibility mapping: a case study of Firooz Abad-Kojour, Iran, Earthquake (2004)

Abstract

Seismic rockfall is one of the prevalent geohazards that cause huge losses in the earthquake-stricken areas. In the present research, a model is developed to map susceptibility (occurrence probability) of seismic rockfalls in a regional scale using Logistic Regression (LR) and Adaptive Neuro-Fuzzy Inference System (ANFIS) techniques. In this research, Firooz Abad-Kojour earthquake of 2004 was introduced as the benchmark and the model base. The susceptible zones predicted by LR and ANFIS methods were compared with the database (distribution map) of seismic rockfalls, by which the results revealed a good overlapping between the susceptible zones predicted by the ANFIS and the field observation of rockfalls triggered by this earthquake. Besides, for the statistical evaluation of results obtained by LR and ANFIS models, the verification parameters with high accuracy such as density ratio (Dr), quality sum (Qs), and receiver-operating characteristic curve (ROC) were used. By analyzing the susceptibility maps and considering the Qs index obtained by LR (21.04184) and ANFIS (26.75592), it could be found that the Qs of ANFIS is higher than that of LR. Moreover, based on the obtained value of the area under the curve (AUC) from LR (0.972) and ANFIS (0.984) methods, ANFIS provided a higher accuracy in zonation and susceptibility mapping of rockfalls triggered by Firooz Abad-Kojour earthquake of 2004 compared to the LR method.