Using the PS-InSAR technique based on amplitude time series analysis for late-appearing persistent targets in potential sliding areas

Potential large-scale landslides and earth-rock flow potential streams are located in remote mountainous areas, most of which overlap with aboriginal tribes. Since the Morakot typhoon disaster, the government has invested many resources in the potential soil and sand disasters. However, the number of such potential areas is too large to fully cover and monitor. After the Wulai District of New Taipei City, Taiwan, was severely damaged by the Sudil typhoon disaster in 2015, many significant landslides happened, and the Central Geological Survey, Taiwan, announced a total of 17 potential large-scale landslides. The area of potentially large-scale landslides in Xiluoan is the largest, about 6.5 km2, and covers the central residential area of the Wulai tribe. 

MT-InSAR (Multi-temporal Interferometric Synthetic Aperture Radar) is a powerful remote sensing technique for ground movement detection. The most popular method is the Persistent Scatterer Interferometry (PS-InSAR) in early 2000. The concept of PS-InSAR is to look for steady, point-like scatterers whose phases are consistent across entire time series SAR data. However, for those scatters are considered temporary targets or late-appearing persistent targets during an entire period; they are often ignored and not estimated. In order to overcome such limitation, an amplitude On-Off model, which uses a rectangular function to estimate temporary targets, was applied to the amplitude time series. Then, we used the On-Off model result as a weight in the InSAR time series processing frames. Two rectangular corner reflectors (CRs) were placed in the Wulai sites, potentially large landslide areas. 83 Sentinel-1 SLC ascending images acquired from Jan 2020 to Dec 2022 were used to estimate ground movement. It is evident that with the weight based on the amplitude on-off model, the late-appearing persistent targets are successfully identified. For example, the two CR locations can be readily found, and the velocity of the movement can be estimated. The amplitude of the pixel at the CR locations becomes significantly strong and stable after the date of setting up the CR targets. One CR covers a period of 20 images; the other covers a period of 31 images. The movement of one CR reveals an ongoing sliding trend. Such estimation is consistent with those typical PS targets located at the same slope sliding area.