The Persistent Scatterer Interferometry (PSI) technique for ground deformation and landslide monitoring using Sentinel-1 images time-series

Abstract

This thesis focuses on the surface motion and landslide monitoring using remote sensing imagery. Landslides are among the most severe and unpreventable geomorphological hazards in the mountainous area. The monitoring and management of ground deformation are necessary for early warning systems. Satellite imageries are effective tools to study the ground deformation. The extraction of information from SAR (Synthetic Aperture Radar) images can provide changes on the Earth’s surface. Displacement maps derived from SAR images show the displacement in the LOS (Line of Sight) direction of the satellite. This technique can provide the capabilities for measuring the deformation in high resolution with a precise accuracy of 1 millimeter/year.

The objectives of this research are to increase the potential of ground deformation monitoring leading to the identification of landslide using PSI and to compare the effectiveness of PSI in urban and rural areas.

This research demonstrates the methodology to overcome limitations of temporal and geometric decorrelation with repeat-pass satellite data using the Persistent Scatterer Interferometry (PSI), which is applied to investigate and monitor the ground deformation. Besides, PSI can increase the potential of ground deformation surveillance leading to the identification of landslides. The trend of ground deformation is obtained by using InSAR time-series analysis of Persistent Scatterer (PS) and Quasi-Persistent Scatterer (QPS) from Sentinel-1 SAR images acquired from 2017 to 2019. The advantages of Sentinel-1 SAR images are the open access and all-weather availability. The PS technique is used to explore stable point-like radar targets, which works well in urban areas. However, it has a limitation to be applied in heavily vegetated areas. The QPS technique is used to analyze in more complex terrain, which can process the most coherence interferogram as a weight for each interferometric phase to improve the coherence target and the spatial distribution of QPS points within vegetated areas. These techniques have significantly applied to monitor the severity of deformation and landslide.

This thesis considers the research questions; How is the ground deformation derived from PSI processing to recognize stable and unstable areas? How effective is the PSI technique to detect ground deformation in urban and rural areas?

To answer these research questions, the Sentinel-1 SAR images were selected. The PSI approach was used to investigate the ground deformation and landslides in both city and rural areas. Beijing, China is capital of China and has been among the most seriously affected city by subsidence. Fangshan District in Beijing was selected due to the occurrence of a massive landslide. Another study area is Nan province located in northern Thailand where landslides occur every year, especially in the rainy season and most of the area in this province is a mountainous area covering with the heavy vegetation.

The results show displacement rates in Beijing city, with -99 to 19 mm/yr. There were two obvious displacements in the eastern part. The susceptible areas of Beijing were related to the geological structure, the density of buildings, and the distribution of infrastructure. One landslide in PS and two landslides in QPS were identified in Beijing rural case study; there were observed as slow-moving landslide with a motion of 5 mm/yr. The displacement map indicated that the QPS was also capabilities of detecting land deformation in Nan case study, with rate of -262 to 248 mm/yr. The most susceptible areas located on hilltops and steep valleys. The QPS results performed is better than the results of PS because they can improve the coherence and the measured points in high and complex terrain.

The displacement maps were generated using PS and QPS; it can be explained in the movement and identification of susceptible areas. The landslides were detectable using both techniques but probably undetectable during continuous monitoring with the PS in heavy vegetation areas. Any landslides loss in detect is a small landslide, located in a valley where no structures as potential good radar targets are reflectance. The results extracted by PS usually have the advantage of high backscatter and stable phase. Thus, the displacement acquired by PS only refers to PS points, but the QPS can be detected the displacement in the area without PS points. The integration of different techniques is significantly advantageous for monitoring land deformation and landslides. However, both techniques will be useful for local development planning and decision making

-