Research On Early Warning Model Of Rainfall Landslide In Loess Landform Combined With InSAR Technology

China is a country with a high incidence of landslide disasters.Landslide disasters account for more than 70% of various geological disasters.Landslide disasters not only cause huge losses to our economy,but also threaten people's lives.As an effective means of preventing landslide disasters,the study of landslide warning has the characteristics of low cost and high feasibility compared with other methods.Factors such as rainfall and ground deformation are the main causes of landslides.The current monitoring methods for rainfall are very mature.Synthetic aperture radar is one of the effective methods for monitoring ground deformation at present.Compared with optical images,it has the advantages of all-day,all-weather,cloudpenetrating fog and high monitoring accuracy.However,the traditional Differential Interferometric Synthetic Aperture Radar(D-InSAR)technique is susceptible to spatiotemporal loss of coherence,atmospheric delay phase,and terrain phase,and has low accuracy in long-term baseline ground subsidence monitoring.In response to the above problems,domestic and foreign scholars have proposed a variety of timing analysis methods,such as Small Baseline Subset(SBAS)time series interferometry and Permanent Scatterer InSAR(PS-InSAR)time series interferometry.Because PS-InSAR is not effective in monitoring large gradient deformation areas,this article starts from rainfall infiltration and combines multiple hazard factors to logistic risk analysis in the study area by logistic regression analysis.Finally,the SBAS technology is used to extract The ground deformation and neural network technology then carried out risk zoning to the study area,and analyzed the spatial relationship between landslide and ground deformation in the study area.The main work and innovations are as follows:(1)This article starts from the relationship between rainfall and different geological infiltration capacity.First,the geological distribution in the study area is determined by the geological distribution map in the study area.Second,the maximum and minimum water content and correlation of various geological types are counted.The parameters,combined with the layered hypothesis infiltration model and the impervious surface data,construct a rainfall infiltration model in the study area.(2)Based on the rainfall infiltration model in the study area,combined with the slope,aspect,vegetation index,impermeable layer and highway data in the study area,and grading it,then collect the modeling points in the study area at Perform logistic regression analysis in SPSS to determine the regression coefficients of each hazard factor,evaluate the relevant parameters of the model to determine the availability of the model,and construct the early warning model of rainfall infiltration loess landslide in the study area through the obtained regression function to obtain the risk zoning map Combined with historical landslide points and field survey landslide points to verify their spatial distribution relationship.(3)On the basis of the above disaster-causing factors,increase the cumulative ground subsidence data obtained by Sentinel-1A data for deformation monitoring in the study area,and add the cumulative ground subsidence as a new hazard factor to the model In the past,the model was trained in conjunction with the BP neural network,and the spatial distribution relationship was verified by field survey of landslide points.