Antonova, Sofia; Sudhaus, Henriette; Strozzi, Tazio; Zwieback, Simon; Kääb, Andreas; Heim, Birgit; Langer, Moritz; Bornemann, Niko; Boike, Julia (2018): TerraSAR-X DInSAR displacement map of Kurungnakh Island (Lena River Delta, Siberia) in summer 2013, link to GeoTIFF file [dataset]. PANGAEA, https://doi.org/10.1594/PANGAEA.894775,

---

In supplement to: Antonova, S et al. (2018): Thaw subsidence of a yedoma landscape in Northern Siberia, measured in situ and estimated from TerraSAR-X interferometry. Remote Sensing, 10(4), 494, https://doi.org/10.3390/rs10040494

Differential SAR interferometry (DInSAR) uses the phase difference between two SAR signals acquired on two dates over the same area to measure small-scale ground motion. During the last decade the method has been adapted for monitoring permafrost-related ground motion. Here we perform DInSAR on TerraSAR-X data to assess its viability for seasonal thaw subsidence detection in a yedoma landscape of the Lena River Delta.

TerraSAR-X is a right-looking SAR satellite launched in 2007, operating in the X-band (wavelength 3.1 cm, frequency 9.6 GHz), with a revisit time of eleven days. All data that we used were acquired in StripMap mode with HH polarization from a descending orbit at 08:34 local acquisition time (22:34 UTC). The incidence angle of the track we use is approximately 31 degrees. The scene size covered an area of approximately 18 km x 56 km. The slant range and azimuth pixel spacing were approximately 0.9 m and 2.4 m, respectively. Based on the ground temperature data we roughly estimated the beginning and the end of thaw season in 2013. The corresponding TerraSAR-X time series used for this study includes nine Single-Look Slant Range Complex (SSC) images taken between 7 June and 14 September 2013. The time span between the acquisitions that we used for interferometry was 11 days, with one exception when the time span was 22 days due to a missing acquisition. The data were processed using the Gamma radar software. The SSC data were converted to Gamma Single Look Complex (SLC) format and the SLC data were then consecutively co-registered with subpixel accuracy (typically better than 0.2 pixels) in such a way that the co-registered slave image became the master for the next image. This way of co-registering also ensures subpixel co-registration accuracy for all interferometric combinations of the nine images. Multilooking was performed with the factor 4 in the range and factor 3 in the azimuth directions to reduce the noise and obtain roughly square ground range pixels. The ground size of the multilooked pixel is approximately 7 m. We removed the topographic phase term using ArcticDEM that is a freely available high-resolution (5 m) circum-Arctic DEM produced from optical stereographic WorldView imagery acquired from 2012–2016. Obtained differential interferograms were then filtered with an adaptive filter based on the local fringe spectrum with the filtering window size of 128 pixels and an alpha exponent of 0.4. Interferograms, featuring especially low coherence, were additionally filtered with a window size of 64 pixels. For the phase unwrapping we used a branch-cut algorithm with the seeding point located approximately in the middle of the study area with relatively high coherence. We did not attempt to unwrap the areas, separated from the main study area by the river channels. The influence of atmospheric phase delays was evident in the unwarpped interferograms. In order to enhance the displacement signal and reduce atmospheric noise, all eight unwrapped interferograms were summed up in a time-continuous stack. Phase rate per day was calculated from the stack. A strong linear ramp was present across the phase rate map. To remove the trend, a 2D linear function was fit to the data and then subtracted from the phase rate map. The phase rate was then converted to vertical displacement rate in meters, under the assumption that the ground movement is purely vertical. The resulting displacement rate map was geocoded using ArcticDEM to the Universal Transverse Mercator (UTM) projection, zone 52N WGS84 with a pixel size of 5 m. The map was finally converted to the displacement magnitude by multiplying the rate by 99 days (from 7 June to 14 September 2013) and converted to centimeters. As opposed to the results, published in the related paper, here we did not start the unwrapping from the known bedrock position, as it was partly affected by low coherence as well as rather remote from the main area of interest and only weakly connected to the rest of the map over a small and noisy area of valid pixels. It means that the displacement map published here, features only displacement values relative to each other, without a fixed reference point. The spatial pattern of the signal, however, did not change with this alteration in processing. The DInSAR map showed a distinct subsidence in most of the thermokarst basins relative to the upland. Moreover, the spatial pattern of DInSAR signal was in high agreement with the surface wetness in the basins, identified with the near infra-red band of a high-resolution optical image. Drier parts of the basins were clearly separated from wetter parts that showed a prominent subsidence. In general, low coherence in combination with atmospheric effects as well as remoteness of a reference ground point were severe obstacles for the retrieval of a wide-area seasonal thaw subsidence map with TerraSAR-X data.