Multiscale Susceptibility Zonation Of Landslides Triggered By Buried Fault Earthquake

As a common part of the earthquake disaster chain,the earthquake landslide not only brought huge loss of life and property,but also seriously affected the earthquake relief work.The physical mechanism of the earthquake landslide is complex and the factors are dificulte to obtain.The susceptibility zonation of landslides triggered by earthquake without ground rupture using GIS and data mining effectively avoids the difficulties of evaluation based on physical model,which makes it an important method to evaluate the susceptibility of earthquake landslides.The Niigata Earthquake in Japan,a buried fault earthquake,was selected as the research object of this paper.The magnitude of the main shock was 6.8,followed by 4 aftershocks with magnitudes greater than 6.0,which triggered thousands of landslides.At the same time,in order to make up for the lack of surface rupture zone,the ground deformation is added as the influencing factor.In this paper,multi-scale regional susceptibility evaluation of seismic landslides based on data mining and GIS is carried out for both of the large area and the epicenter region.The main work and results are as follows:(1)Three data mining models,such as Logistic Regression,Support Vector Machine and Artificial Neural Network,were used to evaluate the susceptibility of the landslide to the large research area.The study area was properly partitioned according to the susceptibility level.The simulation results of the three models were analyzed according to the mathematical evaluation methods such as ROC and the partitioning result of the susceptibility level.The results show that the three models have achieved good results especially artificial neural network.(2)For the epicenter region,the co-seismic surface deformation is added as the new influence factor to the artificial neural network model,which obtained the best effect in the simulation of the large research area.A more detailed seismic zonal susceptibility zoning map is obtained.The results show that the ground deformation has a certain improvement on the results with a greater contribution more than the common impact factors like slope and the distance from the road.Therefore,in the case of insufficient data on influencing factors,ground deformation is valuable especially for earthquakes without surface rupture.(3)The comparison of multi-scale regional evaluation results shows that compared with the evaluation of susceptibility to earthquake landslide in large area,the epicenter area study has obtained a more accurate seismic susceptibility zoning map in the same area,but the area under the ROC curve is commonly smaller.The comparative analysis of the results of multi-scale regional evaluation realizes the multi-scale earthquake prone landslide zoning,which meets multiple levels of earthquake landslide disaster prevention and control needs.