InSAR and SAR Pixel-Offset Time-Series Techniques for Understanding Landslide Movements in the Three Gorges Region.
Singleton, Andrew; Li, Zhenhong; Hoey, Trevor
University of Glasgow, UNITED KINGDOM

The Three Gorges region, China, has long suffered from frequent, widespread and varied landslide hazards. The increased water-level of the new reservoir associated with the Three Gorges Dam Project (TGP) has also reactivated large, ancient landslides. Along with the very practical importance of identifying and monitoring unstable and inhabited slopes, the densely vegetated region provides a useful test for the application of frequently acquired Synthetic Aperture Radar (SAR) images to understand the mechanisms of slow-moving landslides beyond urban areas. Additionally, the characteristics of landslides (rapidly variable rates of movement with sharp boundaries between stable and deforming land) can be major limitations in InSAR analysis that assumes a spatial gradient of deformation to be relatively smooth.

This study compares three different techniques (specifically conventional SAR Interferometry, pixel-offset and pixel-offset time-series) using a vast amount of SAR images from two independent satellite platforms (ESA Envisat and DLR TerraSAR-X).

Coherence analysis suggests there is a strong seasonal component along with rapid temporal decorrelation associated with vegetation growth. Despite this, the highest resolution TerraSAR-X Spotlight data does allow a boundary to be accurately defined around an active slope failure. Through subsequent image pairs a very basic "time-series" of movement can be generated through conventional InSAR techniques and the accumulated deformation over the 15 months of data is consistent with pixel-offset maps generated from a single pair of images as well as a least squared inversion of many pixel-offset maps.

The comparison of techniques helps to validate our measurements without ground data, and the consistency between InSAR and pixel-offset techniques provides confidence in interpreting the horizontal displacement measurements obtainable from the 2-dimensional offset maps. Finally, the time-series of landslide movement is shown to be strongly related with the draw-down of the Three Gorges reservoir to accommodate the summer rainfall. This can also help with interpretations of the failure mechanism that suppose the movement is attributed to a reactivation of an ancient landslide deposit that has been caused by the increased ground water level following the reservoir's completion.

Following recent acknowledgement about the increased landslide hazard after reservoir inundation continued widespread monitoring could be achieved using frequently acquired, high resolution SAR imagery in this challenging terrain, ultimately to help mitigate the negative effects on local populations.