| Earthquake-induced landslide is a common secondary disaster. It causes great harm to the disaster area. The most serious region in Wenchuan earthquake is Sichuan province, due to the characteristics of complex geological conditions, strongly new tectonic activities, rugged terrain, river system rolling and rich precipitation, etc. According to statistics, the number of geological disasters in this area has been the largest comparing with other province of China in previous years. Under the force of the earthquake, the geological disaster in this area featured with the greast number, most complex type and the biggest scale and volume, occurrence of aftershocks combined with heavy rainfall and human activity, the secondary landslide-induced by the earthquake always become more frequently which will restrict the reconstruction work and economic development. Therefore, in the aspects of disaster prevention, disaster mitigation, land planning and disaster management, take researchs on landslide in serious disaster area have practical significance and study value.In this paper, we chose Beichuan County as the study area which is the one of the most secious destroyed region in wenchuan earthquake. This research was taken with the supporting of GIS, the formation mechanism of earthquake-induced landslide and the regional environment charachteristics of the research area. It included two parts, the analysis of the space distribution features of landslides and the evaluation of landslide susceptibility through two different kinds of model. The main research contents and achievements include:(1) Based on the formation mechanism of earthquake-induced landslide, we collected and processed the impact factors data of landslide. According to the change point theory, the optimum statistical radius of windows used to extract relief degree of land surface of the study area was determined in this paper. The statistical correlation between the landslide distribution and the impact factor had also been analyzed. (2) According to the neighboring analysis theory, we discovered that before the earthquake, the landslide points tend to be discrete distributions.While after the earthquake, the landslide points shown gathered distribution pattern and the agglomeration degree was very obvious.This phenomenon indicated that the distribution of earthquake-induced landslide in this area has a significant regularity. The point density was chosen as the index to classifiy the landslide hazard degree and to deeply analyze the characteristics of the three major sever landslide development area. Based on the accumulation and fractal dimension transformation theroy, we draw a conclusion that slope, aspect, undulating terrain, etc affect the landslide seriously, while water, roads, seismic intensity of earthquake, etc are relatively have small impaction on it. These conclusions provide a basis for the selection of subsequent sensitivity evaluation.(3) The multiple regression model and neural network model were applied to landslide spatial prediction and mapping. The model calculated result was ultimately categorized into four classes. They shown that the high and very high susceptibility areas mostly distribute in Qushan, Chenjiaba towns, etc, which along the rivers and the roads around the area of Longmenshan fault. The precision accuracy using multiple regression model is about73.7%, and the neural network model can be up to81.28%.It can be concluded that in this study area, the neural network model appears to be more accurate in landslide spatial prediction. |
PDF Full Text Request