Interferometric synthetic aperture radar data from 2020 for landslides at Barry Arm Fjord, Alaska
Dates
Distribution
2020-07-08
Publication Date
2020-07-07
Revision
2020-07-27
Revision
2020-08-21
Revision
2020-10-22
Last Revision
2020-11-13
Citation
Schaefer, L.N., Coe, J.A., Godt, J.W., and Wolken, G.J., 2020, Interferometric synthetic aperture radar data from 2020 for landslides at Barry Arm Fjord, Alaska (ver. 1.4, November 2020): U.S. Geological Survey data release, https://doi.org/10.5066/P9Z04LNK.
Summary
Subaerial landslides at the head of Barry Arm Fjord in southern Alaska could generate tsunamis (if they rapidly failed into the Fjord) and are therefore a potential threat to people, marine interests, and infrastructure throughout the Prince William Sound region. Knowledge of ongoing landslide movement is essential to understanding the threat posed by the landslides. Because of the landslides' remote location, field-based ground monitoring is challenging. Alternatively, periodic acquisition and interferometric processing of satellite-based synthetic aperture radar data provide an accurate means to remotely monitor landslide movement. Interferometric synthetic aperture radar (InSAR) uses two Synthetic Aperture Radar (SAR) scenes taken [...]
Summary
Subaerial landslides at the head of Barry Arm Fjord in southern Alaska could generate tsunamis (if they rapidly failed into the Fjord) and are therefore a potential threat to people, marine interests, and infrastructure throughout the Prince William Sound region. Knowledge of ongoing landslide movement is essential to understanding the threat posed by the landslides. Because of the landslides' remote location, field-based ground monitoring is challenging. Alternatively, periodic acquisition and interferometric processing of satellite-based synthetic aperture radar data provide an accurate means to remotely monitor landslide movement.
Interferometric synthetic aperture radar (InSAR) uses two Synthetic Aperture Radar (SAR) scenes taken at different times to generate maps of surface deformation in the line of sight of the radar sensor using differences in the phase of returning waves. In addition to ground deformation, changes in phase can occur due to earth curvature, topography, orbital effects, atmospheric conditions, and other noise such as change in ground scattering properties. Removal of phase change contributions other than ground deformation are corrected through InSAR processing, although several limitations exist (e.g. Massonnet and Rabaute, 1993; Gens and Van Genderen, 1996). Phase noise can be estimated using coherence, which is a measure of similarity of the radar path reflection between two SAR images, ranging from 0 (phase is just noise) to 1 (complete absence of phase noise). Interferometric phase as shown in "wrapped" interferograms is the phase of the wave, represented by an angle from 0 to 2π, which returns to 0 after one cycle. Phase unwrapping solves this phase ambiguity by creating a continuous phase, which can be converted to relative displacement in an "unwrapped" interferogram (Yu et al., 2019). The maximum detectable deformation rate is one radar wavelength/2 per pixel. For the RADARSAT-2 satellite data used herein, the SAR wavelength is ~5.6 cm, which means the maximum detectable deformation rate is 2.8 cm within one pixel relative to the next pixel.
Here, we present the interferometric results of tasked RADARSAT-2 SAR data. Data were acquired from two ultrafine beam modes, U19 and U15, that are acquired over the landslide every 24-days, beginning on May 26, 2020 and June 2, 2020, respectively. The spatial resolution is 3 m in range and 3 m in azimuth. Each time period listed below will have an associated unwrapped interferograms (units of line-of-sight displacement in centimeters) in geotiff and kmz format, along with a pdf that will highlight ground features and provide scale and satellite look direction information. For each time period, we describe interferogram results for three areas surrounding Barry Glacier: Landslide A, Landslide B, and the Northwest-facing slope (Fig. 1). Detailed methodology is provided in a separate text file. Results will be added to this data release as new scenes are acquired and processed. Artefacts within the scenes (e.g., due to atmospheric conditions or changing topographic conditions) can be identified by using several independent interferograms, and we will continue to monitor and assess the presence of artefacts over time.
26 May 2020-19 June 2020: This interferogram contains substantial noise, in part due to changes in snow cover over the time period. The very high noise-to-signal ratio results in very few coherent pixels, making interpretation difficult. However, in locations where there are coherent pixels in Landslide A, Landslide B, and the Northwest-facing slope, there is either no landslide movement, or movement is on the millimeter scale. Displacement patterns on the landslide, do not indicate that ground displacements were greater than the upper limit of detection (> 2.8 cm within one pixel relative to the next pixel), which would result in a loss of coherence.
02 June 2020-26 June 2020: This time period has more coherent pixels than the earlier time period. In Landslide A, coherent pixels indicate very little movement within the landslide (typically <1 cm) in the line of site (downslope direction). Although efforts are made to reduce errors, as described in the methodology, this small amount of movement likely falls within the margin of error. Thus, this interferogram indicates that Landslide A is not moving, or is moving on the millimeter scale. Displacement patterns on the landslide do not indicate that ground displacements were greater than the upper limit of detection (> 2.8 cm within one pixel relative to the next pixel), which would result in a loss of coherence. Landslide B shows ~5-8 cm of displacement in the northeast part of the toe of the slide, but the remainder of the slide shows little to no movement. On the Northwest-facing slope, a relative change in phase near the toe of the glacier may indicate small (1-2 cm) movement, and will be further assessed using future interferograms.
19 June 2020-13 July 2020: In Landslide A, coherent pixels again indicate that the landslide did not move during this time interval, or is moving on the millimeter scale. Displacement patterns on the landslide do not indicate that ground displacements were greater than the upper limit of detection (> 2.8 cm within one pixel relative to the next pixel), which would result in a loss of coherence. Landslide B shows up to 6 cm of displacement of the northeast toe of the slide, as noted by reference (1) in the PDF attachement '20200619_20200713.pdf'. Cumulative displacement cannot be determined until further interferograms are processed. The remainder of the slide shows no movement, or movement that is on the millimeter scale. On the Northwest-facing slope, an approximately 0.04 km2 area centered on -148.100, 61.153 indicates up to 2.5 cm movement in the downslope direction, as noted by reference (2) in the PDF attachement '20200619_20200713.pdf'. This movement will be monitored and further assessed using future interferograms. The remainder of the Northwest-facing slope shows no movement, or movement that is on the millimeter scale.
26 June 2020-20 July 2020: On Landslide A, there is a relative phase change on the lower half of the slope and toe of the landslide. These changes indicate that the lower half of Landslide A moved about 1 cm or less, resulting in a slight upward bulge at the toe of about 1.5 cm, as noted by reference (1) in the PDF attachment '20200626_20200720.pdf'. Landslide B shows up to 7 cm of displacement in the northeast part of the toe of the slide, as noted by reference (2) in the PDF attachement '20200626_20200720.pdf'. Thus, between 02 June and 20 July, cumulative downslope movement in this area of Landslide B is at most 15 cm. The remainder of the slide shows no movement, or movement that is on the millimeter scale. On the Northwest-facing slope, an area of approximately 0.09 km2 near the toe of the glacier centered at -148.120, 61.142 indicates up to 3 cm of movement in the downslope direction, as noted by reference (3) in the PDF attachement '20200626_20200720.pdf'. This movement will be monitored and further assessed using future interferograms. The remainder of the Northwest-facing slope either shows no movement, movement on the millimeter scale, or artefacts (e.g. due to changes in snow cover).
13 July 2020-6 August 2020: In Landslide A, coherent pixels again indicate that the landslide did not move during this time interval, or is moving on the millimeter scale. Landslide B shows ~2-7 cm of displacement of the northeast toe of the slide, as noted by reference (1) in the PDF attachment '20200713_20200806.pdf'. Thus, between 19 June and 6 August, cumulative downslope movement in this area of Landslide B is ~13 cm in the areas with the largest deformation. On the Northwest-facing slope, an approximately 0.04 km2 area centered on -148.100, 61.153 (WGS84) indicates up to 1.3 cm movement in the downslope direction, as noted by reference (2) in the PDF attachment '20200713_20200806.pdf'. The remainder of the Northwest-facing slope either shows no movement, movement on the millimeter scale, or artefacts (e.g. due to changes in snow cover).
20 July 2020-13 August 2020: Due to a large satellite orbital drift between acquisitions (perpendicular baseline = 251.57 m), this interferogram has a decrease in coherence and several artefacts related to our inability to correct phase distortions in steep terrain at this time. On Landslide A, there is deformation on the upper half of the slope, as noted by reference (1) in the PDF attachment '20200720_20200813.pdf'. This suggests 1.5 cm of movement or less, which may partly be due to orbital drift errors and will be assessed in future interferograms. Reference (2) shows apparent deformation at the toe of the landslide, which is the opposite of the change seen in the previous interferogram of this beam mode (26 June 2020-20 July 2020), and is larger. This is also the case for reference (3) on the Northwest-facing slope and in the area south of the Coxe Glacier. Because of these discrepancies, and the lack of any similar signal in 13 July 2020-6 August 2020 interferogram, which covers a similar time frame but has a much smaller satellite orbital drift (perpendicular baseline = 16.12 m), these apparent movements are likely errors associated with differences in the satellite viewing geometries. These areas will be monitored and further assessed using future interferograms. On Landslide B, where movement was previously identified at the northeast toe of the slide, the interferogram is incoherent, and thus accurate deformation measurements cannot be made in this area from this interferogram. Near the headscarp of Landslide B, there is an area that suggests 2-3 cm of downslope movement as noted by reference (4) in the PDF attachment '20200720_20200813.pdf'. This area will be monitored and further assessed using future interferograms.
06 August 2020-30 August 2020: Due to a large satellite orbital drift between acquisitions (perpendicular baseline = 99 m), this interferogram likely has several artefacts related to our inability to correct phase distortions in steep terrain. These apparent movements are noted by references (1) and (2) in the PDF attachment '20200806_20200830.pdf'. In Landslide A, coherent pixels indicate that the majority of the landslide did not move during this time interval, or is moving on the millimeter scale. The area noted by reference (1) indicates up to 2 cm of movement at the toe of the landslide, but may be an artefact due to differences in satellite viewing geometries. On Landslide B, where movement was previously identified at the northeast toe of the slide, the interferogram is incoherent, and thus accurate deformation measurements cannot be made in this area from this interferogram. There is also apparent movement along the Northwest-facing slope noted by (2) in the PDF attachment '20200806_20200830.pdf', however this pattern of continuous deformation is only apparent in interferograms with large satellite orbital drift between acquisitions, thus this deformation is likely an artefact due to satellite viewing geometries. Location (3) on the Northwest-facing slope shows up to 3 cm of movement in an area where movement
was previously identified, but the magnitude of deformation may be exaggerated due to satellite orbital drift.
13 August 2020-06 September 2020: In Landslide A, coherent pixels indicate that the majority of the landslide did not move during this time interval, or is moving on the millimeter scale. A new area of movement noted by reference (1) in the PDF attachment '20200813_20200906.pdf' indicates up to 1.5 cm of deformation in a small area near the northeast toe of the landslide. On Landslide B, where movement was previously identified at the northeast toe of the slide, the interferogram is incoherent, and thus accurate deformation measurements cannot be made in this area from this interferogram. On the Northwest-facing slope, an area of westward movement indicated by (2) in the PDF attachment '20200813_20200906.pdf' shows up to 1.5 cm of movement in an area where movement was previously identified. Other apparent movement of less than 1 cm indicated by (3) along the entire Northwest-facing slope may be an artefact due to satellite viewing geometry.
30 August 2020-23 September 2020: Due to a large satellite orbital drift between acquisitions (perpendicular baseline = 200 m), this interferogram has very low coherence, and likely has several artefacts related to our inability to correct phase distortions in steep terrain at this time. Landslide B has very few coherent pixels. Although coherent pixels within Landslide B are pink and red (indicating movement), these also extend outside of the landslide and onto the Barry Arm glacier, which we would not expect to have the same deformation pattern. Therefore, it is difficult to differentiate deformation vs. baseline or unwrapping errors in this timeframe in Landslide B. Apparent movement up to 3 cm on the toe of Landslide A, as noted by reference (1) in the PDF attachment '20200830_20200923.pdf', and apparent movement up to 6 cm of the Northwest slope, as noted by reference (2) in the PDF attachment '20200830_20200923.pdf', are both areas that show movements in other interferograms with large perpendicular baselines (e.g., interferogram between 20 July 2020 – 13 August 2020 with a perpendicular baseline of 251.57 m). Thus, these apparent movements are likely related to errors associated with differences in satellite viewing geometries.
6 September 2020-30 September 2020: This scene has similar features to the previous scene from this beam mode (13 August 2020-06 September 2020). In Landslide A, coherent pixels indicate that the majority of the landslide did not move during this time interval, or is moving on the millimeter scale. An area of movement noted by reference (1) in the PDF attachment '20200906_20200930.pdf' indicates up to 2 cm of deformation in a small area near the northeast toe of the landslide, for a maximum of 3.5 cm of movement between the 13 August and 30 September. On Landslide B, areas of movement previously identified at the northeast toe of the slide are incoherent, and coherent pixels within the rest of the landslide indicate no movement or millimeter-scale movement. This is also the case on the Northwest-facing slope.
23 September 2020-17 October 2020: In Landslide A, coherent pixels indicate that there was movement during this time frame. Discontinuous coherent pixels prevents the extraction of data from the entire slide, but several areas of movement are apparent in both the wrapped and unwrapped interferograms shown in the PDF attachment ‘20200923_20201017.pdf’. During the unwrapping process, discontinuous and irregular ground movement can lead to unwrapping errors that result in a more smoothed characterization of movement than what actually occurred. Therefore, to best characterize movement, both the wrapped and unwrapped interferograms are included for this time period. The wrapped interferogram shows the phase of the synthetic aperture radar wave, where each “fringe”, or one full color cycle (i.e. purple to purple), correlates to 2.8 cm of movement. References (1)-(4) in the PDF attachment ‘20200923_20201017.pdf’ are the same for both the wrapped and unwrapped interferograms, but the following details refer to the wrapped interferogram. In the central part of the slide, as noted by reference (1), approximately 6 fringes indicate about 17 cm (6 x 2.8 cm) of movement. In the northern part of the slide, as noted by reference (2), fringes are irregular and potentially show several different zones of movement, with about 5 fringes indicating 14 cm of movement in the area nearest to the head scarp. As noted by reference (3), an area near the south end of landslide also experienced movement, with about 1.5 fringes indicating 4.2 cm of movement. Landslide B is largely incoherent, and thus accurate deformation measurements cannot be made in this area. On a small portion of the Northwest slope, as noted by reference (4), one fringe may indicate 2.8 cm of movement.
30 September 2020-24 October 2020: This scene contains substantial noise, likely from atmospheric conditions and snow fall during the time frame. The very high noise-to-signal ratio yields very few coherent pixels, resulting in insufficient measurements and therefore no viable interpretation of the scene.
Summary of movement during the 2020 monitoring period: This data release has described the deformation of two landslides and the northwest-facing slope at the Barry Arm Fjord between 26 May and 24 October, 2020. Early (26 May– 19 June) and late (30 September – 24 October) scenes contained substantial noise, likely because of changes in snow cover or other atmospheric conditions, which resulted in very few coherent pixels and thus insufficient measurements. Additionally, we found that using the 5 m Interferometric Synthetic Aperture Radar (IFSAR) Alaska Digital Terrain Model (DTM) from 2010 to create unwrapped interferograms did not fully remove topographic phase, especially in scenes with large satellite drift between acquisition dates. This problem was likely a result of changes in the landslide topography between 2010 and 2020 (Dai et al., 2020). All movement described is in the line of sight (LOS) of the satellite, where positive values indicate movement away from the satellite and/or subsidence, and negative values indicate movement towards the satellite and/or uplift in the unwrapped interferograms. A small area near the northeast toe of Landslide A had apparent movement of 3.5 cm between 13 August – 30 September. Otherwise, between 2 June – 30 September, the majority of Landslide A either showed no movement or moved on the millimeter scale. Between 30 September – 17 October, the central portion of Landslide A moved about 17 cm downslope. Although the data is discontinuous, there is evidence that several areas throughout the landslide experienced movement. Between 2 June – 6 August, the northernmost part of the toe of Landslide B moved about 15 cm downslope. After 6 August, the parts of the interferograms covering Landslide B were largely incoherent and accurate deformation measurements were not possible. The Northwest-facing slope had two areas of movement over this time frame. An area of approximately 0.04 km2 centered on -148.100, 61.153 (WGS84), moved about 7 cm downslope between 19 June and 6 September. An area of approximately 0.09 km2 near the toe of the glacier centered at -148.120, 61.142 (WGS84) moved about 5 cm downslope between 26 June and 17 October. All cumulative movement mentioned in this paragraph was determined by summing movement from individual scenes.
References cited:
Dai, C., Higman, B., Lynett, P.J., Jacquemart, M., Howat, I.M., Liljedahl, A.K., Dufresne, A., Freymueller, J.T., Geertsema, M., Ward Jones, M. and Haeussler, P.J., 2020, Detection and assessment of a large and potentially‐tsunamigenic periglacial landslide in Barry Arm, Alaska: Geophysical Research Letters, 47(22), e2020GL089800. https://doi.org/10.1029/2020GL089800.
Gens, R., Van Genderen, J. L., 1996, Review Article SAR interferometry—issues, techniques, applications. International Journal of Remote Sensing, 17(10), 1803-1835.
Massonnet, D. and Rabaute, T., 1993, Radar interferometry: limits and potential. IEEE Transactions on Geoscience and Remote Sensing, 31(2), 455-464.
Yu, H., Lan, Y., Yuan, Z., Xu, J., & Lee, H., 2019, Phase unwrapping in InSAR: A review. IEEE Geoscience and Remote Sensing Magazine, 7(1), 40-58.
Any use of trade, firm, or product names is for descriptive purposes only and does not imply endorsement by the U.S. Government.
We present the interferometric results of tasked RADARSAT-2 SAR data from 2020 at Barry Arm Fjord in southern Alaska. Data were acquired from two ultrafine beam modes, U19 and U15, that are acquired over the landslide every 24-days, beginning on May 26, 2020 and June 2, 2020, respectively. Each time period will have an associated unwrapped interferogram (units of line-of-sight displacement in centimeters) in geotiff and kmz format, along with a pdf that will highlight ground features and provide scale and satellite look direction information. For each time period, we describe interferogram results for three areas surrounding Barry Glacier: Landslide A, Landslide B, and the Northwest-facing slope. Results will be added to this data release as new scenes are acquired and processed. Artefacts within the scenes (e.g., due to atmospheric conditions or changing topographic conditions) can be identified by using several independent interferograms, and we will continue to monitor and assess the presence of artefacts over time.
Revision 1.1 on July 27, 2020. Revision 1.2 on August 21, 2020. Revision 1.3 on October 22, 2020. Revision 1.4 on November 13, 2020. To review the changes that were made, see "VersionHistoryReadme.txt" in the attached files section