Persistent Scatter Interferometry (PSI) processing

Persistent Scatter (PS) Interferometry (PSI) is a powerful remote sensing technique able to measure and monitor displacements of the Earth’s surface over time. PSI represents a specific class of DInSAR (Differential Interferometric Synthetic Aperture Radar) techniques, which exploits multiple SAR images acquired over the same area. It utilizes appropriate data processing and analysis procedures.

The processing starts with the Single Look Complex (SLC) images, the precise orbits of the sensor and an external Digital Elevation Model (DEM) of the area. It mainly consists in three steps:

• Interferogram stack generation, which includes image co-registration, interferometric filtering (if necessary) and differential interferogram generation. The latter is obtained by cancelling out the topographic term of the interferometric phase with the DEM and the precise orbits.

• Pixel selection, different approaches are available if the interest is in deterministic, distributed scatters; or both.

  • Dispersion of Amplitude

  • Coherence

  • Temporal Sublook Coherence

  • Temporal Phase Coherence

• Deformation estimation, which basically consists on the adjustment of a linear model to data, to obtain the linear velocity of deformation and DEM error, and the application of filters in both time and space, to retrieve the time-series of deformation.

The main outcomes of PSI techniques are the ground deformation time series and the precise geolocation of the pixel.

Another outcome of a PSI analysis is the so-called residual topographic error or DEM error, which is the difference between the true height of the scattering phase center of a given PS and the height of the DEM in this point.

CommSensLab has developed SUBSIDENCE-GUI a user friendly interface for PSI processing that implements the Coherence Pixels Technique (CPT).

Screenshot of SUBSIDENCE-GUI

The software is being commercialized by DARES (http://dares.tech).

PSI ground deformation monitoring result

Mexico City subsidence (26 May 2017 to 21 May 2018) PSI monitoring result by 30 Sentinel-1B VV SAR images. Negative values mean moving away from the satellite, i.e. subsidence in this case. The maximum subsidence reaches up to around -25 cm/year.

Mexico City is sinking and in the last hundred years it has sunk more than 10 meters. This obviously poses a grave problem for this megacity and its 21.1 million inhabitants that has been caused by the groundwater extraction. The land subsidence has caused millions of dollars’ worth of damage to infrastructures such as buildings, water pipes and sewer lines, subway tunnels and roads and clearly requires a continuous monitoring to prevent further consequences and, if possible, define remedial strategies.

Landslide monitoring results

Canillo landslide motion retrieved from 32 Staring Spotlight TerraSAR-X (TSX) Single Look Complex (SLC) SAR images with a 0,23m resolution in azimuth and 0,59m in range. Deformations have been projected to the down-slope direction.

The study has demonstrated that Super High Resolution (SHR) data, jointly with advanced pixel selection strategies, can dramatically improve the number of high-quality pixels for its later PSI processing, which is of crucial importance in landslide monitoring in natural environments. The high density of PS obtained allows the achievement of a more robust network of PS (improving the linear estimation without propagation errors and the reliable estimation of APS) and thus favors the reliable estimation of displacement maps in a major number of points inside a landslide. The three main subareas with noticeable displacement of the landslide have been detected, which are similar to those obtained in previous PSI monitoring results. The PSI measured displacement rates have been compared with GPS measurements of the same period, and they are both in good agreement.

To investigate the temporal evolution of the Canillo landslide, the down-slope time-series displacement results obtained by the TPC method at two different PSs have been plotted. The displacements observed for both PSs are exhibiting considerable nonlinear components, presenting some acceleration and deceleration periods within each year. From the two PSs’ 2016 displacement time-series, we can find that the stable periods start at the beginning of July and end at the middle of August. These periods are coincident with the trend of Canillo averaged monthly precipitation, where the lowest precipitation is in July with an average of 79 mm, as Figure 13e shows. This indicates that the movements of the landslide have some seasonal patterns, which are correlated with the amount of precipitation.

PSI DEM errors estimation result

Besides the displacement results, PSI techniques can also obtain the DEM error of the selected pixels with respect to the reference DEM used. The inclusion of the retrieved DEM error on the geocoding of the final results largely improves the geolocation quality of the displacement maps.

The next figure shows some interesting examples that illustrate the capabilities of SHR TSX data to retrieve the vertical distribution of scatters in manmade structures. The examples shown have been obtained from the TPC processing. Subfigure (a) shows a communications tower located in Canillo. The vertical distribution of scatters perfectly follows the tower’s structure as the picture validates. It is also interesting, looking at the GoogleEarth image, to compare the distribution of scatters with the shadow of the tower projected over ground. Subfigures (b) and (c) show a couple of chairlifts from the Grandvalira ski station. Once again, the vertical distribution of scatters perfectly follows the metallic structure, as the pictures and projected shadows demonstrate. Finally, subfigure (d) shows a couple of high voltage towers. The good performance of the vertical location of the scatters, thanks to the inclusion of the calculated DEM error on the geocoding process, can also be used as proof of the reliability of the displacement velocity maps obtained. Both velocity and DEM error have been calculated simultaneously when adjusting the linear model to the interferometric data.