Shallow lakes (less than 3-4 m deep) are a widespread feature in many regions underlain by permafrost (e.g. Alaska, Canada and Russia). There has been interest for many years in monitoring/mapping the seasonal evolution of floating and ground-fast ice (i.e. ice that is frozen to the bottom of the lake) in shallow lakes from SAR remote sensing. Knowing when (i.e. the timing) and where the ice becomes ground-fast or remains afloat during the ice growth season is relevant for climate monitoring. With Arctic climate warming, shallow lakes are showing signs of thinner late-winter (April-May) ice covers in some regions (e.g. North Slope of Alaska), resulting in a smaller fraction of lakes that become ground-fast. From a permafrost perspective, a shift from a ground-fast ice to a floating-ice regime can initiate talik (thaw bulb) development underneath lakes that could potentially release large stocks of carbon previously frozen in permafrost in the form of methane.

As a contribution to ESA’s DUE GlobPermafrost project (2016-2019), algorithms for the generation of a ground-fast lake ice product from Sentinel-1A/B synthetic aperture radar (SAR) imagery were implemented, cross-compared, and validated. Sample pixels (200 pixels per image) were randomly selected from a set of 30 Sentinel-1 SAR images (five images for each of six lake regions from Alaska, Canada and Russia). The images used for accuracy assessment were all acquired at/near the time of maximum ice thickness (mid-April to early May). Ground-fast ice and floating ice visually determined from the sample pixels by an ice analysis were compared with output from three different algorithms (Thresholding, Iteration Region Growing with Semantics (IRGS) and K-means). The thresholding algorithm (92.4%) was found to perform slightly better on average than the IRGS algorithm (90.1%), and to significantly outperform K-means (85.3%). The thresholding algorithm was therefore selected for implementation of a fully-automated processing chain to generate a novel ground-fast lake ice product. Using dense time series of Sentinel-1 A/B SAR imagery acquired in Extra Wide Swath (EW) and Interferometric Wide Swath (IW) modes, the new map product consists of the day of year (DOY) when the ice becomes ground-fast. We will present these recent results in view of the development of an operational system for Arctic-wide ground-fast lake ice mapping from historical, current and upcoming SAR satellite missions.

More than one third of the Earth‘s coastlines are comprised by Arctic permafrost coasts. Erosion rates of up to 25 m yr^-1 at specific locations and retreat rates of for example 0.5 m yr^-1 along the Yukon coast in Northwestern Canada indicate a high vulnerability towards Climate change. This leads to alterations in biochemistry, a redistribution of soil organic carbon (SOC), contaminants, sediment and other nutrients in arctic permafrost regions. In the context of coastal erosion, a release of e.g. carbon, soil organic matter and contaminants into the ocean is likely. In addition, lateral fluxes from catchments into the ocean change under altered climatic conditions.
As part of the EU project Nunataryuk our goal is to map and characterize landcover and landforms at catchment scale using high spatial resolution optical imagery in combination with Synthetic Aperture Radar (SAR). Landforms in the catchments are portrayed by high- and low-centered ice wedge polygons at different stages of degradation. In consideration of their distinctive shape, Object-Based Image Analysis will be used for classification. In contrast to pixel-based methods, neighborhood-relations are taken into account which is in particular relevant for landcover dominated by a distinctive shape.
Further, Machine Learning Digital Soil Mapping (ML DSM) methods will be applied to model the spatial distribution of different soil parameters, in particular SOC stocks. For training and prediction of the model, environmental variables derived from satellite imagery and SAR will be used. Previous work has shown that the application of Machine Learning (ML) methods can gain promising results, particularly the method Random Forest. In contrast to geostatistical methods, that require high point density in the research area, ML DSM methods offer the opportunity to improve data particularly in regions with difficult and costly accessibility. It is further possible to receive results at a high spatial resolution. Small scale studies are currently less widespread and our goal is to determine variances in SOC in a small catchment. These results will contribute to the determination of lateral fluxes of Dissolved Organic Carbon into the ocean.

Rock glaciers are sediment bodies consisting of a mixture of rock debris and ice typical of mountain permafrost. Rock glaciers have a distinct morphology indicating flow like down slope movement by deformation of the ice contained within them. Annual rates of motion are ranging from some millimetres to a few metres per year, varying within the annual cycle, from year to year, as well as at the decennial time scale. During the last decade, an increasing number of studies have monitored in-situ the creep behaviour of rock glaciers in the European Alps. Observations showed that most rock glaciers surveyed, irrespective of size and velocity, responded sensitively and almost synchronously to inter-annual and decennial ground temperature changes. Whereas the absolute motion rate of a rock glacier is depending on a multitude of factors, relative variations over time (from months to years) appear to be in particular depending on the temperature profile between the permafrost table and the main shear horizon at depth. Changes in rock glacier motion are thus believed to be the most indicative short- to medium-term response of rock glaciers to environmental changes and an indicator of mountain permafrost conditions in general. Besides being environmental indicators, rock glaciers and their changes impact on geohazards and the water cycle in high mountains.

Satellite radar interferometry (InSAR) has been efficiently used in the past for the estimation of the rate of motion of rock glaciers. Previous works were mainly devoted to the inventory of rock glaciers and other periglacial processes in combination to optical data, providing outlines and state of activity (or class of motion) of the mapped landforms. The irregular acquisitions and the long acquisition time intervals of past SAR missions prevented the long-term monitoring of rock glacier kinematics. With the Sentinel-1 constellation, specifically designed for SAR interferometric applications and displacement monitoring, SAR images of the same orbit are regularly acquired every 6 days over Europe and every 12 days over other mountainous areas worldwide. The mission is operational since 2014 with an expected lifetime of 20 years and images are distributed following a free and open data policy. In our work we analyse the performance of Sentinel-1 InSAR time-series for the monitoring of the rate of motion of rock glaciers. Representative rock glaciers in the Swiss Alps, Disko Island (Greenland), and the semi-arid Andes of sufficiently large extend and moving at rates of motion of a few millimetres to a couple of centimetres in 6 days are considered. The line-of-sight displacement in 6 to 48 days derived from the interferometric pairs is converted to rates of motion along the slope. Sentinel-1 based rock glacier velocity time-series show in many cases an annual cycle, with velocities higher at the end of the summer (Figure 1). Reliable InSAR information can be derived for the snow-free season and in winter during stable, cold conditions. A quantitative assessment with in-situ methods will be also presented.

In periglacial landscapes, the seasonal ground freeze and thaw induce heave and subsidence due to the phase change of the water within the ground. On slopes, mass-wasting processes contribute to the development of various creeping landforms. Changes of the ground thermal regime in a context of climate change can lead to modification in the spatial distribution, the amplitude and timing of frost-related ground displacements. In this context, it is paramount to develop measurement techniques able to monitor periglacial environments on a regional scale, and to explore the relations between ground displacements and environmental variables.

Synthetic Aperture Radar Interferometry (InSAR) based on Copernicus Sentinel-1 datasets allows for millimetric measurements of ground displacements over large and remote areas. Our research focuses on two study areas in complementary periglacial contexts: Adventdalen and the adjacent valleys in Nordenskiöld Land (Central Svalbard) with continuous permafrost; and Rásttigáisa-Geaidnogáisá area in Finnmark (Northern Norway), with sporadic-discontinuous permafrost. In Svalbard, we compared two-dimensional InSAR results from combined ascending and descending SAR geometries to a geomorphological map. We show that InSAR allows for the identification of specific displacement signatures for different typical periglacial materials and landforms. Using Small BAseline Subets (SBAS) technique, seasonal thaw subsidence and frost heave patterns were detected in June to December 2017 time series with 6-days sampling. The displacement patterns match the ground thermal dynamics of the active layer measured in boreholes (Rouyet et al., submitted).

In order to further understand the relations between displacements and environmental variables controlling the ground dynamics, InSAR results can valuably be exploited for the modeling of spatial-temporal frost-related displacements. In our study, we investigate the potential of combining InSAR maps and time series with physical-based (Hu et al., 2018; Liu, et al., 2018; Westermann et al., 2017) and statistically-based (Hjort & Luoto, 2013) modeling techniques. Temporally, InSAR subsidence and heave patterns are compared with modeled time series, using the Stefan equation based on thawing and freezing degree-day indices and more advanced transient ground thermal models. Spatially, statistically-based explanatory modeling allows for studying the complex relations between InSAR displacements and a set of climatic, topographical and geomorphological variables.

References:
- Hjort, J. & Luoto, M. (2013). Statistical methods for Geomorphic Distribution Modeling. In: Shroder, J. (Editor in Chief), Baas, A.C.W (ed.), Treatise on Geomorphology. Academic Press, San Diego, Vol.2. Quantitative modeling of geomorphology, 59– 73. DOI:10.1016/B978-0-12-374739-6.00028-2.
- Hu, Y., Liu, L., Larson, K.M., Schaefer, K.M., Zhang, J., Yao, Y. (2018). GPS Interferometric Reflectometry Reveals Cyclic Elevation Changes in Thaw and Freezing Seasons in a Permafrost Area (Barrow, Alaska). Geohpysical Research letters, 45(11), 5581– 5589. DOI:10.1029/2018GL077960
- Liu, L., Rouyet, L., Strozzi, T., Lauknes, T.R., Christiansen, H.H. (2018). Seasonal Thaw Settlement and Frost Heave in Permafrost Regions in the Arctic: A Synthesis of InSAR Observations Using Sentinel-1 SAR Images. Accepted abstract. American Geophysical Union Fall Meeting, 10–14 December 2018, Washington D.C, U.S.A.
- Rouyet, L., Lauknes, T.R., Christiansen, H.H., Strand, S.M., Larsen, Y. (2018, submitted). Seasonal Dynamics of a Permafrost Landscape, Adventdalen, Svalbard, investigated by InSAR. Manuscript submitted for publication.
- Westermann, S., Peter, M., Langer, M., Schwamborn, G., Schirrmeister, L., Etzelmüller, B., Boike, J. 2017. Transient modeling of the ground thermal conditions using satellite data in the Lena River delta, Siberia. The Crysophere, 11, 1441– 1463. DOI:10.5194/tc-11-1441-2017.

Thermokarst development in ice-rich permafrost regions is a natural hazard, and causes irreversible geomorphological changes. Thermokarst is the process by which characteristic landforms result from the thawing of ice-rich permafrost or the melting of massive ice. The formation of large depressions in the ground surface produced by thermokarst processes results in surface inundation, and causes damages on infrastructure. Eastern Siberia is one of the regions where ice-rich permafrost is broadly distributed and thermokarst has been advancing quickly. The remarkable surface subsidence due to thermokarst has been observed in recent years, which could cause destruction of building and infrastructure in this area (Shiklomanov et al., 2016). However, little quantitative observation is available to examine where and how the surface deforms. Remote sensing technique, especially Interferometric Synthetic Aperture Radar (InSAR), has a possibility to monitor the thermokarst subsidence and seasonal surface displacement over the permafrost regions. Recent studies of permafrost monitoring using SAR data have been reported, and SAR could provide essential information to understand thawing process of ice-rich permafrost and to mediate resulting destruction of infrastructure.
In this study, we used ALOS/PALSAR (2007-2011) and ALOS-2/PALSAR-2 (2014-2018) data to investigate ground subsidence caused by thermokarst development. GAMMA software was used to generate Single Look Complex data from Lv1.0 data in ALOS/PALSAR and Lv1.1 data in ALOS-2/PALSAR-2. AW3D, 5 m digital surface model derived from ALOS/PRISM, was used to simulate and remove topographic fringe. We generated some interferograms and applied stacking procedure weighted on period between two SAR data acquisition period. Assuming ground displacement is only vertical component, the LOS change was converted to vertical displacement using incidence angle.
In the Mayya area, on the right bank of the Lena River close to Yakutsk, we detected ground subsidence with a rate of 1-4 cm/yr in both PALSAR and PALSAR-2 results. The subsidence signals are found in numerous open areas (deforested areas), and the PALSAR-2 results clearly show the spatial distribution of the subsidence corresponding to visible observation of thermokarst development in high-resolution optical images. They also roughly coincide with the in-situ observation in 2017-2018. Based on the site visit in September 2018, the two large subsidence signals in the PALSAR-2 results are found in alasses. Alas is considered as a final geomorphological stage of old thermokarst development. Significant amount of the subsidence may be due to consolidation settlement associated with soil desiccation under recent dry climate conditions. However, judging from new developments of polygonal subsidence in the bottom areas of alasses, ground ice still remains under the alasses, which may cause further thermokarst progress.

Deep-seated gravitational slope deformations as well as permafrost creep (i.e. rock glaciers) represent common geomorphological phenomena in high alpine areas. The systematic observation of changes over time caused by the slope movements is of high importance for a proactive management of the natural hazards related to these phenomena.
Ground-based dynamic monitoring, such as Global Positioning System (GPS) or total station, presents the disadvantage of a limited spatial coverage and is difficult to conduct in remote or impervious areas. These limitations become even more evident for monitoring dynamics at subseasonal scale.
Satellite remote sensing methods based on Synthetic Aperture Radar (SAR) data represent a set of complementary technologies at disposal to study displacement of rock glaciers and landslides at different space and time scales. In fact, SAR satellites are capable to acquire data over large distributed remote areas relatively independent from weather conditions and with high temporal frequency (up to a few days).
In this work, we propose a SAR-based approach that aims at supporting in-situ measurements for well instrumented study areas, and at monitoring rock glaciers and landslides dynamics when in-situ measurements are not available or difficult to be carried out.
The primary test site of this study is the Lazaun rock glacier in the Schnals valley. It presents displacement velocities up to 1.5 m/year, and although GPS measurement campaigns have been conducted there since 2006, it has never been monitored before by using remote sensing techniques.
SAR data with different characteristics (i.e. different spatial and temporal resolution) were processed using different techniques, i.e. differential interferometry (DInSAR), multi-temporal interferometry and amplitude tracking, to study the annual, seasonal and subseasonal rock glacier dynamics with high precision.
In detail, the high acquisition frequency of the Copernicus Sentinel-1 data (6 days) allowed studying the subseasonal displacement and understanding the displacement rate variation during the snow free period over the test area selected. Three datasets of Sentinel 1 from relative orbits 15, 117 and 168 were used, taking into account the images from the snow free period 2017 (about mid-June to November). The images were processed using Persistent Scatter (PS) and Small BAseline Subset (SBAS) techniques. Interferometric techniques allowed obtaining only the displacement along the Line of Sight (LOS) of the satellite. Exploiting the combination of the three relative orbits, displacements in three dimensions can be derived.
Six Very High Resolution (VHR) TerraSAR-X (TSX) images acquired between summer 2016 and summer 2018 were also processed by using differential interferometry (DIn-SAR). In order to overcome the limitations of DIn-SAR techniques related to high displacement rates, decorrelation and unwrapping issues, amplitude tracking technique was applied on VHR TerraSAR-X images. By exploiting the combination of the displacement along the LOS and along the satellite flight direction, displacement direction is estimated
The accuracy of satellite-based products is assessed exploiting ground-based and proximal data, such as GPS, Ground-Based SAR (GB-SAR) and UAV-based data (acquired concurrently with the GPS campaigns). Three GB-SAR campaigns were carried out during summer 2017 and 2018. The high acquisition frequency (5 minutes) of this instrument enabled the comparison of daily displacement data with satellite SAR results.
The second test site is the Corvara mass movement, classified as a deep seated complex landslide with rotational (i.e. upper part) and flow-like (i.e. middle and lower part) behaviour. The moving volume of this slide is about 30 million m³ affecting a surface area of 2.5 km² and has been under observation since 1997. The slide can be subdivided into source, track and accumulation zones and is characterized by spatially varying displacement rates. The presence of vegetation induces a quite strong decorrelation effect in SAR. As a cconsequence in order to allow for monitoring the area using these techniques, a set of corner reflectors equipped with a support for GNSS-antenna were installed. The corner reflectors are characterized by two different size designed for X and C band respectively. The availability of periodic GPS measurements, permanent GPS stations, UAV flight campaigns, Sentinel1 and 3 TerraSAR-X data allowed to test amplitude tracking and multi-interferometric techniques in this area in a similar way with respect to Lazaun test site.

Acknowledgments:
This work was conducted within the project ALPSMOTION (ALPine Slow slope Movement moniTorIng and detectiON with remote and proximal sensing), coordinated by Eurac Research-Institute for Earth Observation and funded by the Autonomous province of Bolzano, Alto Adige, “Ripartizione Diritto allo Studio, Università e Ricerca Scientifica.” The Copernicus Sentinel-1 and TSX data were processed with SARscape software (SARMAP). GB-SAR instrument has been made available by University of Pavia. GPS data were provided by University of Innsbruck. TSX Data were provided by the European Space Agency, Project Proposal id 34722, © DLR, distribution Airbus DS Geo GmbH, all rights reserved.

Figure caption: Example of Din-SAR results using TerraSAR-X data over the Lazaun rock glacier (images acquired on July 14 and August 8, 2017