Study On Risk Assessment Technology Of Regional Seismic Landslide

Seismic landslide natural disaster because of its particularly large extent of the damage became widely studied issue in recent years. With the continuous development of 3S technology, study of earthquake landslide is developed more deeply, which is chaning gradually from the previous qualitative analysis to quantitative analysis phase, and this is becausemodern remote sensing technology can provide us with massive data which is from different angles, levels, and depths. Also scientists have developed many precise physical and mathematical models, which can improve the reliability of estimates of landslide.By processing Wenchuan landslideson GISplatform,Landslide layersare converted into a raster grid and expressed in different grade.Landslide Landslide samples are generated in the center of gravity; while sampleswithout slipping is generated in the area where the landslide didi not occur. Based on the evidence weight model, the two classification logic regression and BP neural network mathematical models, Icarry out the risk assessment of landslide hazardfor the study area(E103 ° ~ E105 °, N30.8 ° ~ N32 °).According to the weight of the evidence, the weight of each influence factor is obtained.In other word, we can gain the probability of the impact factor triggering the landslide. Finally, by using the spatial analysis of GIS, each impact factor layer will be combined together to get the seismic landslide hazard zoning map of the study area.Evaluation curve showed a steep-convex trend and the area under the curve accounted for 90.24%.Using SPSS software and implementing multiple logistic regression analysis on 17,286 sample records, we could get the contribution-weights which representthe possibility of triggering-landslides and by which we could carry out overlay-lay analysis to obtain Seismic landslide hazard assessment map. According to natural-classification method, the evaluation map will be divided into very-low, low, medium, high, very-high five levels regions, and 70% of the landslide occurred in the predicted very-high areas. The evaluation of seismic landslide convex curve showing a steep trend, It showed that the using of multivariate logistic regression is feasible to evaluate the seismic landslide.This article used BP neural network methodto identify the earthquake and landslide-prone evaluation studies. The results show: BP neural network landslide recognition correct rate reached 85.3%, and 70% of the landslide occurred in the predicted high-risk areas, and the evaluation of seismic landslide convex curve showing a steep trend, It showed that the using of BP neural network is feasible to evaluate the seismic landslide.By applying the three particle swarm optimization neural network algorithm to the identification of Wenchuan landslide,I compared the non-optimized results with three differentoptimized neural network.The results show that the optimized neural network convergence speed is accelerated apparently. The correct rate of recognition and evaluation of the proportion of the area under the curve rate increased by about 3 percentage points.On the basis of the C# visual programming environment and the component Arc Engine, I design the Geographic information system for landslide hazard assessment of regional earthquake, which is used to evaluate the seismic landslide hazard zone during the earthquake occurrence, and provides support for the rescue and rescue of earthquake landslide.