The Early Identification Of Potential Landslides In Loess Plateau Based On Time-series In Sar And Terrain Analysis

Loess Plateau Area is featured with special geography and geology environment,where landslides and other geological disasters are prone to occur by trigger factors like rainfall and human activities and then cause significant loss of life and property to local people.Suide,as an important town in the Loess Plateau,has frequent landslide hazards.Taking the Suide county town as a case for surface deformation monitoring and potential landslides identificationis of significance for monitoring and early identification of urban geological disasters in the Loess Plateau.Meanwhile,it has practical significance for the prevention and control of geological hazards in the Suide county.The research area of this paper is Suide county town and its surrounding area,covering an area of 60 square kilometers.Using DInSAR technology to obtain surface deformation before and after the 726 flood event in the study area in 2017;using change detection of image technology to obtain surface horizontal displacements from2016.12.05 to 2017.12.10.The PS-point deformation process is divided into linear and nonlinear types by time series analysis.The nonlinear type deformation curve is compared with the landslide displacement-time curve,selecting the type(mutant type)that suits to the landslide deformation process as a feature,and identifing the slope unit.The slope,terrain diversity index,horizontal displacement that obtained by image change detection technology,and deformation speed acquired by SBAS technology were used as raw data.The potential landslides obtained from field surveys are used as training samples to construct a neural network model,and with it to identify potential landslides in the study area.The conclusions were obtained as follows:(1)Based on the optimization of filter parameters and the topology relationships of image connection etc.,PSInSAR,SBAS,and DIn SAR technologies are used to obtainsurface deformation in the direction of line-of sight.By comparison,the deformation results obtained by different means are comparable.There is a correlation between the line-of-sight direction displacement that obtained by PSInSAR and the horizontal displacement that obtained by image change detection technology,which shows that the deformation data is reliable and can reflect the surface deformation process in the area.(2)In the study area,there is a general trend of subsidence.The subsidence rate of the urban area is relatively large.The larger deformed areas of the surrounding mountainous areas show scattered distribution,and the PS-points with the same deformation are obviously concentrated in the local area.There is a good correspondence relationship between rainfall changes and deformation results obtained by PSInSAR.While verifying the surface deformation results are reliable,it also reflects that the main factor of surface deformation is precipitation.(3)Through the time series analysis of the deformation process of the PS-point,the result show that the mutation type is in line with the process of landslide deformation.The break point of the mutation type mostly occurs in June to July,according with the rainfall time.The PS-point of mutation type subsides before and after the flood event.The trend of Subsidence is obvious,indicating that the mutation type has traceability for identifing potential landslide.(4)The neural network model was used to combine the deformation features with the terrain features for identifing potential landslide.That results were in good according with the field survey results.The identified landslides are discretely distributed,which is consistent with the fact that the stable structural and rainfall is the main cause of landslides.The above results show that the landslide identification method used in this paper is effective and more effective than the potential landslide identification based on a single feature.