Localized landslide risk assessments with multi pass X, L band DInSAR and pixel tracing analyses

Monday, 15 December 2014
Hyewon Yun, Jungrack Kim, YunSoo Choi and Kisu Kim, University of Seoul, Seoul, South Korea

The anthropogenic construction activities frequently occur in the cutting side of mountainous area and result in localized landslide associating with multiple environmental factors. Although Differential Interferometric SAR (DInSAR) analysis is attractive tool to foresee the susceptibility of landslide, it is difficult to attain sufficient enough accuracy in DInSAR measurement due to many error factors such as the vegetation canopy which causes the decorrelation of InSAR phase, the highly steep slope and distribution of water vapor inducing the propagation delay.

In this study, we tackled such problems employing the solutions as follows; i) DInSAR analysis with L band ALOS PLASAR producing higher correlation in weakly vegetated area, ii) the water vapor observation from weather Research Forecasting (WRF) with 1 km spatial resolution, iii) the usage of high resolution LIDAR DEM or additional algorithm to compensate the correlated error between baselines and DEM offsets.

The target areas of this study were chosen in the eastern part of Korean peninsula centered in Uljin and Gangneung respectively. In there, the landslide originated by the geomorphic factors such as high sloped topography and localized torrential down pour is critical issue. The landslide susceptibilities in Uljin are crossly compared between the results using L band DInSAR analysis and GIS interpretation of landslide triggering factors such as vegetation, slope and geological properties. Gangneung was targeted by multi pass X band SAR images. Thus, sub meter resolution pixel tracing and DInSAR analysis were combined to decompose precise displacement vectors which are more useful to be compared with the ground based deformation measurement. After then the local landslide hazard map together with the GIS interpretation of landslide triggering factors was extracted.

The study will be further extended for the application of future SAR sensors by incorporating the dynamic analysis of topography to implement practical landslide forecasting scheme.