Conference Agenda

Single-pass across-track (XTI) SAR interferometry (InSAR) has been widely used to measure changes in glacier volume and ice sheet topography. A main concern is the elevation bias resulting from the penetration of the radar signal into the snow/ice [1]. SARlab at Simon Fraser University (SFU) is operating a Tri-Band airborne SAR system (X, C & L Band – dual, single and quad-pol respectively), co-mounted with an optical system for structure from motion (SfM) photogrammetry [2]. The optical system is being operated in oblique looking configuration for maximum swath overlap with the SAR sensor to serve multiple purposes in terms of enhanced motion compensation and providing high resolution snow surface DEMs for the area of interest. The SFU airborne X and C band SAR sub-systems are being operated in across-track (XTI) single pass InSAR configuration, while the L Band system is being operated in along-track (ATI) configuration. The field area for the snow penetration experiment in Kluane National Park Reserve (KNPR), Yukon Territories, Canada consists of non-polar icefields with snow firn and glaciers, all within close vicinity to our system base at the Silver City airfield. The optical derived DEM can generate 10 cm resolution DEMs as a snow surface topography reference for our XTI configuration, which allows precise snow depth penetration measurements at both X and C band frequencies. The experiment’s main objective is to gain better understanding of how snow penetration at C-band is influenced by incidence angle and snow properties varying with elevation on the icefield and its outlet glaciers as this is relevant to correct firn and snow elevation biases of ESA mission Harmony, who funds this experiment.

Algorithms are being tested on preliminary datasets collected over firn area with our system in September 2022. A dedicated campaign is scheduled for the experiment in mid-April 2023, where we will measure the snow penetration for three sites at different elevation, relative to the optical surface DEM and using exposed rock in the SAR swath (identified in the optical ortho mosaic) as surface phase reference for the InSAR DEMs. We plan to fly several partially overlapping swaths at each site to attain incidence angle diversity and to strengthen the optical SfM solution. At one of these sites, we validate the accuracy of our “exposed rock-referencing method” for the single pass InSAR phase with two metal corner reflectors (CR) installed on the surface to provide an additional alternate phase reference. Ground truthing of snowpack properties (as a combination of snowpit and core drilling to five meters) will be carried out at the sites, as close in time as possible, to allow proper interpretation of the acquired data. A preliminary analysis of our results will be shown at the conference, including how accurate single pass coherence can be used to estimate the snow penetration bias at C-band with the method of [3]

Keywords: Singlepass InSAR, across-track interferometry, snow depth penetration, snow penetration phase bias

References

[1] Dall, Jrgen. "InSAR elevation bias caused by penetration into uniform volumes." IEEE Transactions on Geoscience and remote sensing 45.7 (2007): 2319-2324.

[2] J. Stacey, W. Gronnemose, J. Eppler and B. Rabus, "En Route to Operational Repeat-Pass InSAR with SFU’s SAR-Optical Airborne System," EUSAR 2022; 14th European Conference on Synthetic Aperture Radar, Leipzig, Germany, 2022, pp. 1-5.

[3] Dall, Jørgen. (2007). InSAR Elevation Bias Caused by Penetration into Uniform Volumes. Geoscience and Remote Sensing, IEEE Transactions on. 45. 2319 - 2324. 10.1109/TGRS.2007.896613.

The Sentinel-1 (S1) satellites have been gathering valuable Extra Wide swath (EW) data over the polar regions since their launch in 2014, with primary applications in maritime operations, oil spill detection, and sea ice monitoring. While the technical design of the EW mode is similar to the Interferometric Wide (IW) swath mode, most of the data in the EW archive is only available as level-0 or Ground Range Detected (GRD), requiring the ability to focus raw level-0 data to Single Look Complex (SLC) data to use interferometric SAR techniques (InSAR). With the support of ESA, Norwegian Research Centre (NORCE) has developed this capability, enabling novel research on terrestrial applications within the cryosphere. In this presentation, we will showcase results from an ESA pilot study (EW-EXPLORE) where we investigate the potential of the EW data archive with InSAR techniques to push boundaries in research on ice shelf dynamics in East Antarctica and permafrost dynamics in the Arctic archipelago of Svalbard.

Ice shelves, formed by land ice that enters the ocean and starts floating, are of particular importance to the stability of the ice sheet and are found distributed along the Antarctica coast, coincidently where most of EW data is acquired. By using 3- and 4-pass double differential InSAR we managed to create time series (2016-2021) of grounding zone observations, providing the first grounding line estimates over a major ice shelf in East Antarctica since 1994 and thus enabling a long-term change assessment. We will also discuss strengths and weaknesses of EW to study ice surface velocities and ice shelf crack propagation, where the dense coverage of EW in space and time enables opportunities for detailed sub-annual monitoring.

In Svalbard, EW is used to retrieve seasonal displacement time series due to the ground ice formation and melting in the active layer above the permafrost, similarly to what has been done with conventional IW mode. We selected snow-free seasons in 2020 and 2021, processed the results with a Small Baseline Subset (SBAS) algorithm and found contrasting subsidence/heave amplitudes and patterns driven by inter-annual climatic variability. The EW-based displacement patterns are well comparable with equivalent results based on IW-mode images. In addition, thanks to the high number of overlapping tracks at this latitude, the interferograms from 6 descending tracks with 1 day of temporal shift have been generated and can be combined to provide a comprehensive displacement time series with a daily resolution.

Our results highlight the added value of S1 EW to complement and extent beyond existing InSAR products based on S1 IW mode and improve our understanding of the terrestrial cryosphere in remote regions. Although the project focuses on two specific domains –glaciology in East Antarctica and permafrost science in Svalbard– we conclude that Sentinel-1 EW data has considerable potential to be exploited to an even larger range of applications than originally intended.

The amount of water contained within a snow pack is the Snow Water Equivalent (SWE), which is an important parameter for climate and hydrological models. SWE estimates are needed to make accurate flood predictions in the snow melt season and are important for water resource planning and management. However, SWE in-situ measurements can only be made on a limited number of locations and are especially challenging in many snow-covered areas due to low accessibility. A wider coverage can be obtained using remotely sensed data. Synthetic Aperture Radar (SAR) can monitor large areas and is independent from weather and illumination conditions. Another advantage is, depending on the frequency, the ability of radar waves to penetrate into the snow pack and being therefore sensitive to snow properties, like depth, density, anisotropy and SWE.

A powerful tool for mapping SWE is Differential Interferometric SAR (DInSAR). Since the dielectric constant of snow differs from the one of air, radar waves are refracted in the snow pack. This has an influence on the optical path length of the radar wave. If the SWE has changed between two acquisitions, the difference in path length can be measured with the interferometric phase [1], [2], offering high potential for SWE monitoring. However, one limitation of the method is that the interferometric phase lies in the interval [-π, π], leading to a phase wrap, when the SWE change exceeds a frequency dependent threshold.

For this study, different SAR data sets are used. Ground measurements can be used to detect the amount of phase wraps. By adding the missing phase cycles inferred from the ground measurements, the DInSAR SWE retrieval results can be corrected and, thus, significantly improved. Due to the limited availability of ground measurements, a SWE parameter derived from a meteorological model, that is parametrized for the region of interest, will be utilized to detect the phase wraps over a larger area.

Another way to estimate the amount of phase wraps is polarimetric SAR. The Co-polar Phase Difference (CPD) can be calculated between the VV and HH polarized channel and correlates with the amount of fresh snow. With the model from [3], the CPD can be inverted to the fresh snow depth. The potential of including polarimetric variables into the DInSAR SWE retrieval algorithm to obtain a more accurate SWE estimation is investigated and compared to the ground measurements and modeled SWE data. First results indicate that polarimetry provides snow depth information and can therefore help to estimate the amount of phase wraps in the DInSAR phase. By correcting these phase wraps in the retrieval algorithm, a higher agreement between the estimations and ground measurements is achieved.

In this study the different ways of correcting the phase wraps are presented, compared and quantitatively evaluated.

[1] T. Guneriussen et al., "InSAR for estimation of changes in snow water equivalent of dry snow," IEEE Trans Geosci Remote Sens, vol. 39, no. 10, pp. 2101-2108, Oct. 2001.

[2] S. Leinss et al., "Snow Water Equivalent of Dry Snow Measured by Differential Interferometry," IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens., vol. 8, no. 8, pp. 3773-3790, Aug. 2015.

[3] S. Leinss et al., “Anisotropy of seasonal snow measured by polarimetric phase differences in radar time series,” The Cryosphere, vol. 10, no. 4, pp. 1771–1797, Aug. 2016.

Rock glaciers are widespread in European Alps and significant for their content of Alpine permafrost. Indeed, they are characterised by a mix of ice and rock, which is related to the presence of permafrost in mountainous areas. The landslide-like behavior of rock glacier is a complex mechanism influenced by the interaction of several factors such as topographical predisposition, internal structure, debris granulometry, temperature, hydrology, and stress conditions. The external temperature is considered one of the most important factors controlling rock glacier flow variation at both inter-annual and seasonal time scales, showing mean velocities ranging from centimetres to meters per year. Hence, the temperature rising due to climate change leads to changes in kinematics of rock glaciers that increase hazards for mountainous settlements and infrastructures.

Despite differential SAR interferometry (DInSAR) is a very effective tool for measuring ground stability, its application to rock glacier monitoring poses several critical issues. First, the steep topography may lead to unfavorable illuminating conditions in terms of either unfeasible detection over layover and shadow areas, or low sensitivity to the ground displacement. Second, the presence of dense vegetation and changeable snow cover conditions causes DInSAR signal decorrelation. Third, displacement kinematics are characterised by both linear and non-linear components and high displacement rates leading to measurements often corrupted by aliasing. This work investigates the rock glacier stability in Val Senales (Italian Alps) by exploiting both the interferometric phase and amplitude of SAR image stack at C-band and X-band.

A multi-temporal DInSAR processing of 345 Sentinel-1 SAR images acquired between 2015 and 2022 was performed by exploiting both persistent and distributed scatterers through SPINUA algorithm. Ad hoc processing strategies were adopted in order to overcome both signal decorrelation due to changeable snow cover conditions, and aliasing due to very high displacement rates. The algorithm was run by selecting spring-summer acquisitions, and forced to search for solutions corresponding to phase changes behind the aliasing limit. The resulting mean line of sight (LOS) displacement map show several areas affected by ground displacements, which lay on exactly within the borders of rock glaciers derived from inventory maps. In some cases, a lack of DInSAR coherent targes occurs just within rock glacier borders, being possibly caused by very high displacement rates not properly measured by the MTInSAR algorithm despite ad hoc processing. These areas were further investigated by exploring maps of DInSAR phase and coherence generated from consecutive Sentinel-1 acquisitions, as well as changes occurring in orthoimages from different years.

Moreover, in order to overcome the DInSAR limitations related to high deformation rates, offset tracking techniques were experimented, which exploit SAR amplitude instead of phase. This analysis was focused on the interesting case study of Lazaun rock glacier [1]. It is a tongue-shaped, 660 m long and 200 m wide, active rock glacier located in Senales Valley (Italy) at about 2600 m asl. Interannual and seasonal displacement rates up to few mm/day are reported by previous studies, which used different techniques including GNSS, inclinometers, and both ground based and spaceborne SAR systems. Offset tracking algorithms can be used to measure displacements with a sensitivity that is a fraction of the data spatial resolution. For the Lazaun case study, we adopted the intensity tracking algorithm, considering that the alternative algorithm based on coherence tracking, is unfeasible due to the low coherence values encountered in the test area. Considering the topography, the size of the area of interest, and the expected entity of the displacement, SAR data acquired along ascending orbits in spotlight mode are those more reliable for displacement estimation through intensity tracking. In particular, we selected six TerraSAR-X staring spotlight and six COSMO-SkyMed Second Generation (CSG), both with a pixel spacing of less than 1m, acquired in the snow free period between 2016 and 2018 (TerraSAR-X) and in 2022 (CSG). These datasets were processed by optimizing the parameters according to the characteristics of Lazaun test case. The displacement maps derived along azimuth and range directions allowed to investigate both seasonal and inter-annual movements occurring on the rock glacier. GPS field campaigns were also carried out in correspondence with some of the satellite acquisitions. A comparison of the results obtained with ground and satellite data were performed showing for the annual displacement a root mean square difference of 0.347 and 0.355 mm/day, with a Pearson coefficient of 0.883 and 0.895 in azimuth and range direction respectively. These results coming from offset tracking provide useful displacement information within the Lazaun borders, where the MTInSAR approach instead suffer of lack of coherent targets due to phase aliasing.

Finally, both mean rates and displacement time series were ingested into a GIS environment together with other informative layers such as multi-temporal mean SAR amplitude, DInSAR coherence maps, rock glacier classes (according to [2]), optical orthoimages, permafrost index map, and Difference Vegetation Index (NDVI). Then, the rock glacier activity was reclassified by adopting the more recent procedure proposed in [3], which is based also on the DInSAR products. This new classification was compared to that derived according to [2] showing several differences. For instance, 3 out of the 6 rock glaciers classified as indefinite were reclassified as relict or translational, 6 out of the 11 rock glaciers classified as relict were reclassified as transitional, and conversely, one rock glacier classified as active was reclassified as relict.

References

[1] C. Fey and K. Krainer, “Analyses of UAV and GNSS based flow velocity variations of the rock glacier Lazaun (Ötztal Alps, South Tyrol, Italy),” Geomorphology, Vol. 365, 2020, 107261. https://doi.org/10.1016/j.geomorph.2020.107261.

[2] E. Bollmann, L. Rieg, L., M. Spross, R. Sailer, k. Bucher, M. Maukisch, M. Monreal, A. Zischg, V. Mair, K. Lang, and J. Stötter, “Blockgletscherkataster in Südtirol-Erstellung und Analyse,” Permafrost in Südtirol, Innsbrucker Geographische Studien. J. Stötter & R. Sailer Eds., pp. 147–171, 2012.

[3] IPA Action Group - Rock glacier inventories and kinematics. Towards standard guidelines for inventorying rock glaciers: practical concepts (version 2.0), pp. 1–10, 2022.

Acknowledgments

This work was carried out in the framework of the project “CRIOSAR: Applicazioni SAR multifrequenza alla criosfera”, funded by ASI under grant agreement n. ASI N. 2021-12-U.0. TerraSAR-X data were provided by the European Space Agency, Project Proposal id 34722, © DLR, distribution Airbus DS Geo GmbH, all rights reserved.

Despite the importance of the seasonal snow cover as a key component of the water cycle, the current observing systems are not able providing adequate, area-wide information on the mass of snow (the snow water equivalent, SWE). Repeat-pass differential SAR interferometry (RP-InSAR) offers a well-defined, physically based approach for mapping SWE at high spatial resolution by measuring the path delay of the radar signal propagating through a dry snowpack. By now the method has not been applied for routine applications, on one hand because of the lack of regular acquisition of suitable RP-InSAR data for covering snowfall events of different intensity, on the other hand due to the need for elaborating procedures towards optimum SWE products covering different types of snowfall events and environments. We report on experimental studies towards the development of consolidated procedures for routine application of the RP-InSAR method in SWE monitoring. The activities comprise field experiments at different sites and the analysis of airborne and satellite-based C- and L-band SAR data, the radar frequencies suitable for applying the RP-InSAR method to retrieve SWE.

There are various critical issues we addressed in these experiments. The RP-InSAR phase does not provide an absolute measurement of the change in snow mass (Delta-SWE) during the time span covered by the interferogram but contains unknown offsets. In order to obtain SWE values, a reference phase is needed for each contiguous coherent area, referring to points with known changes of SWE (e.g. at recording snow stations) or snow-free sites. Other issues are the need to account for penetration losses in vegetated areas (open forests, etc.) and correct for changes in atmospheric phase delay. The latter can be compensated by using the phase of the reference points and/or by using numerical meteorological data on atmospheric water vapour content. The main limiting factor for routine application is temporal decorrelation caused by changes in the complex backscatter signal due the snowfall.

We report on results of field campaigns and on the evaluation of satellite data, addressing these issues. In March 2021 an experimental airborne campaign was carried out in the high Alpine test site Woergetal near Innsbruck. The activities were carried out by DLR HR and ENVEO within the ESA project SARSimHT-NG. Multiple C- and L-band SAR data were acquired by the airborne F-SAR system on 7 days between 2 and 19 March 2021, spanning two snow fall events of different intensity, with mean SWE accumulation amounting to 15 mm and 65 mm. The data analysis focused on impacts of snowfall on the coherence and the performance of retrieved SWE products of the two frequencies, using as reference in situ measurements in different sections of the test site. For SWE retrieval we applied the conventional RP-InSAR method and the delta-k method applying split bandwidth interferometric processing. The mean RP-InSAR Delta-SWE biases of the different tracks are within ±1.5 mm for event 1 and ±6 mm for event 2 (L-band). Delta-k enlarges the measurement range for SWE well beyond the 2p phase ambiguity of conventional InSAR, but has lower sensitivity in respect to changes in SWE. For example, the amount of the SWE changes of the second snowfall exceeds about two-fold the C-band 2 PI phase ambiguity. The C-band data of the 2^nd snowfall event show a largely reduced sample of coherent pixels, but still a sufficient number of valid phase values for the delta-k retrieval.

Further activities were concerned with the development and evaluation of tools and products for SWE retrieval based on satellite data, using L-band data of PALSAR-2 and SAOCOM and C-band data of Sentinel-1. In support of these activities, we performed field measurements in the Upper Engadin, Switzerland, throughout one winter season. In extension of the activities in Alpine test sites, we studied also the performance of C- and L-band InSAR SWE retrievals and potential for area-wide application in areas of the Artic tundra zone.

The studies confirm the importance of L-band RP-InSAR data as basic tool for comprehensive, spatially detailed SWE monitoring. Whereas in L-band the coherence is preserved over extended periods also in case of intense snowfall (as long as it is dry), the C-band coherence degrades strongly during snowfall events of moderate and high intensity. The use of C-band, providing high sensitivity in respect to changes in SWE, will focus on detecting and mapping snowfall amounts of low intensity. Continuous RP-InSAR time series are essential for delivering SWE throughout a winter season by adding up SWE changes of the different sequential periods. If the time series is interrupted, the delta-k method would provide an option for bridging gaps, as well as for capturing extreme snowfall that exceed the range of conventional RP-InSAR products. Furthermore, delta-k interferograms can support phase unwrapping in order to link discontiguous areas.