Present Situation And Spatial Prediction Of Collapse And Landslide In Shilou-ji County Section Of The Middle Reaches Of The Yellow River

The geological conditions of the middle reaches of the Yellow River are complex,and the geological disasters of collapse and landslide are extremely developed.The analysis of the characteristics of collapse and landslide geological hazards in this area and the spatial prediction can not only be effectively applied to the disaster prevention and reduction project,but also provide examples and scientific basis for the subsequent study of geological hazards.However,there are few studies on the characteristics of collapse and landslide and the applicability of spatial prediction based on different mathematical statistical models.In view of this,this paper takes the shilou-jixian section of the middle reaches of the Yellow River as an example to analyze the characteristics of geological hazards and make spatial prediction from the distribution and development characteristics.Were selected in 174 of the 175 history ZaiHaiDian the collapse and landslide ZaiHaiDian as the training sample,with the help of ArcGIS platform sensitivity of collapse,landslide disaster evaluation: first by ArcGIS extract slope,slope direction,elevation and drainage effect distance,the road distance,stratigraphic lithology,rainfall,land use type,vegetation coverage,soil erosion intensity and residents to influence from a total of 11 kinds of influence factors,and then use SPSS software to carry on the principal component analysis and chi-square conditional independence test,nine meet the conditional independence indexes respectively.Combined with the local geological conditions,the indicators are subdivided into a number of secondary factors.Respectively evidence model,information model and build the logistic regression model,calculate the index weights(evidence model,information model)and the regression coefficient,using ArcGIS platform,grid calculation and analysis,and get a posteriori probability values O,total information value I and probability value P,USES the natural breakpoint method,the sensitivity level is divided into: high,high,medium,low,very low.At last,58 historical collapse sites and 59 landslide disaster sites were used as verification points,and the accuracy of these two methods was analyzed by using the proportion of collapse sites and landslide sites in the extremely sensitive and highly sensitive areas and the characteristic curve ROC of the tested workers.Through the above research,the paper has obtained the following results and conclusions:(1)The collapses in this area are all of medium and small size,and are dominated by soil collapses,nearly three times as many as rock collapses.Rock collapses are mainly distributed in the area of hangaozu mountain in jixian county,jihu line in guantou mountain and the inner side of 309 old national road,the urban area of daning county and xiangshuiwan in yonghe county.(2)The landslide scale in this area is mainly small and medium-sized,all of which are soil landslides.It can be divided into four expansion modes: traction type,passage type,expansion type and reduction type.Among them,there are 124 tractive landslides at most,followed by the elapse landslide.The landslide movement forms include lateral extension,compound,flow,translation and rotation.Among them,82 are compound ones,followed by translation ones.There are four planar forms of landslide: semicircle,tongue,irregular and rectangle.(3)Three models were constructed for the sensitivity evaluation of the collapse disaster,and the proportion of the collapse points in the verification area in the high and extremely sensitive area was statistically verified.It was found that the value of 51.72% in the evidential weight model was higher than that of the other two models.Through the ROC curve,the AUC values of the evidential weight model,the information content model and the logistic regression model are 93.1%,83.2% and 90.1%,respectively.These two methods prove that the accuracy of the evidential weight model is the highest in the spatial predictionof the collapse disaster in this area.(4)Three models were constructed for sensitivity evaluation of landslide hazard,and the proportion of landslide points in the high and extremely sensitive areas was statistically verified.It was found that the logistic regression model was 68.97%,significantly higher than the other two models.The AUC values of the evidential weight model,the information content model and the logistic regression model obtained by the ROC curve were 85.5%,85.3% and 93.6%,respectively,which proved that the logistic regression model had the highest accuracy in the spatial predictionof landslide disaster in this region.