Measuring Large-gradient Landslide Displacements And Predicting The Time-of-failure With Spaceborne SAR Remote Sensing

The concealment and destructiveness of landslide disasters pose big threats to the safety of people's lives and property.The surface displacement is vital for evaluating the stability of a landslide and can be a precursory signal of the potential slope failure.Compared with groundbased measurements and spaceborne/airborne optical remote sensing,spaceborne SAR remote sensing has its unique advantages in measuring ground displacements,including high accuracy,large coverage,and all-weather,all-time capabilities.SAR remote sensing has demonstrated its feasibility as an efficient tool for identifying potentially unstable slope and monitoring the stability of landslides.Phase-based SAR interferometry(In SAR)and time-series In SAR are only suitable for mapping slow-moving landslides with displacement rates ranging from several mm/yr to a few tens cm/yr,while amplitude-based pixel offset tracking(POT)is better for detecting and monitoring fast-moving landslides with a large gradient which are beyond the ability of InSAR.The accuracy of POT measurements depends on the spatial resolution of SAR image.The traditional POT method has some problems,such as low measurement accuracy and insufficient of reliability.It is essential to find a way to improve the accuracy and reliability in measuring displacement of large-gradient landslide.Three-dimensional(3D)surface displacements can describe the real movement of slope surface,which is of great significance for analyzing the structural and kinematic parameters of the landslide.Serious image distortions due to SAR viewing geometry in mountainous area usually hinder 3D displacement retrieval from two SAR datasets of different orbits.It is necessary to study how to retrieve the 3D displacements of a landslide when only ascending or descending data is available in mountainous area.In addition,the feasibility of using spaceborne SAR data to predict the time-of-failure of potential landslide collapse disaster also need to be discussed.In view of the problems above,the main researches and contributions in this study are addressed as follow:(1)An adaptive pixel offset tracking(APOT)method was adopted to avoid the matching failure near the landslide boundary.In this paper,an adaptive matching window was adopted instead of the rectangular window in the traditional POT method,thus avoiding the problem that the sliding pixels and stable pixels are inevitably mixed in the same matching window near the boundary and improving the reliability of the deformation measurements.The applicability of APOT method is demonstrated by analyzing the post-failure event of the Huangnibazi landslide.(2)A time-series POT algorithm(TS-PTOT)based on point targets was proposed to increase the accuracy and reliability of POT method.We introduced the idea of Small BAseline Subset(SBAS)in time series In SAR into the point-like target offset tracking(PTOT)method.The accuracy and reliability of deformation measurement using TS-PTOT method were improved compared with that of the single master method.TS-PTOT and TS-POT methods were applied to measure the time-series displacements of two fast-moving landslides with large deformation gradient,the Guobu landslide and the Baige landslide.(3)Two solutions were presented for 3D displacements retrieval in the situation where only images in ascending or descending orbit are available.We applied these two methods to measure 3D displacements of the Guobu landslide and the Baige landslide respectively.The method that combines two descending SAR orbits with different look angle is applied to Guobu landslide,while the method jointly uses SAR and optical data is adopted in Baige landslide.The retrieved 3D displacements help analyze each landslide.(4)The possibility of forecasting the landslide failure using SAR data was demonstrated in the retrospective analysis of two failure events.We successfully applied the inverse velocity method to the retrospective prediction of the time-of-failure of the Baige landslide and the Xinmo landslide.We have demonstrated that the systematic operation of spaceborne SAR can help predict the approximate time of landslide failure and may help with early warning of landslide disasters.