Increasing ground temperatures are expected to have serious consequences for ecosystems, hydrological systems, and infrastructure integrity in the Arctic [1
]. A network of in situ monitoring sites exists but many regions are underrepresented [2
]. Satellite measurements can only indirectly support permafrost monitoring [3
]. Apart from ground temperature, the variation of the seasonal thaw depth on top of permafrost has been addressed with various remote sensing methods including land cover as proxy and displacement monitoring. The latter relies on Synthetic Aperture Radar interferometry (InSAR). Displacement in line of sight can capture the subtle changes (millimeter to centimeter) of ground subsidence and heave, (e.g., [4
]). These seasonal variations in ground displacements are expected to reflect changes in thaw depth and active-layer thickness (ALT) respectively in permafrost areas [5
]. The surface changes result from melt of pore ice or ice lenses. Drainage may also lead to secular subsidence. The water/ice content of a soil therefore determines the amount of displacement. If the soil profile is uniform and is fully saturated, ALT can be directly inferred [7
]. Models which reflect the varying soil layer properties and variations in wetness (precipitation) need to be used in other cases, (e.g., [8
]). An example for an account of different soil layers is described in Wang et al. [8
] and Zhao et al. [9
] (also includes the impact of precipitation). This does, however, require detailed knowledge of the soil stratigraphy, which is rarely available across the Arctic. A further shortcoming is the lack of in situ measurements of active-layer thickness. Such measurements are often sporadic when related to expeditions and therefore do not allow detailed interpretation of findings from InSAR analyses, (e.g., [10
]). Measurement grids of long-term monitoring sites which are part of the Circumpolar Active-Layer Monitoring (CALM) network provide very valuable data in this context. In situ measurements of subsidence itself are also challenging. No in situ monitoring network for ground displacements in permafrost areas exists to date.
Satellite data of all common frequencies (X-, C-, and L-band) have been shown to be applicable to detect line of sight (LOS) displacements related to thaw subsidence. Constraints caused by the wavelength exist for the comparably short X-band (~3.1 cm) due to its sensitivity to changes on the ground between acquisitions [11
] (e.g., for high latitude tundra). This is especially the case if acquisitions come from a single satellite constellation only [12
], what results in case of TerraSAR-X in 11 day repeat intervals. An alternative provides for example COSMO-Skymed (COnstellation of small Satellites for Mediterranean basin Observation), which currently consists of four satellites and allows for repeat measurements of one day. Such data have been so far not investigated for permafrost related applications. Longer wavelengths (specifically L-band, wavelength 23 cm) are of advantage for connecting the seasonal time series across the years [4
]. All so far available long-term application studies make therefore use of L-band data, specifically the Japanese ALOS (Advanced Land Observing Satellite)-missions, (e.g., [5
]). It has been recently demonstrated that also C-band (wavelength 5.6 cm) can be used to establish long-term records in areas with limited vegetation growth [14
]. Large parts of Arctic permafrost areas are, however, characterized with tundra shrubs.
Relating observations between seasons allows also to account for data gaps at the beginning of the soil thaw season. Data pairs usually start several days to weeks later due to the disturbance of snowmelt for repeat InSAR retrieval [16
] and mission specific repeat intervals. This limits the monitoring based on separate seasons, (e.g., [17
]). The majority of ground thaw in areas with unsaturated soils does typically occur right after snowmelt, (e.g., [18
]). A solution would be a constellation as available by COSMO-Skymed. Such data are, however, scarce compared to the acquisitions by C-band missions such as Sentinel-1, which only feature repeat intervals of 6 or 12 days. The latter is more common due to global acquisition strategies. A method to adjust for the early season data gap is required to allow the use of Sentinel-1 or similar data in the extensive region of shrub tundra and unsaturated soils. We hypothesize that knowledge gained from a multi-satellite constellation such as COSMO-Skymed, which provides dense time series, could lead to a suitable approach to account for the early thaw data gap in time series from Sentinel-1. This requires comparability of the displacement measurements from the two different wavelengths. So far only a few studies are available for comparisons between X and C-band over permafrost terrain, (e.g., [4
]). These efforts have been limited to qualitative assessment due to the differences in acquisition timing between the missions. Mostly spatial patterns are assessed visually. A comparison of L-band and X-band demonstrated the impact of vegetation on applicability of TerraSAR-X with comparably long revisit intervals [20
We have selected a long-term monitoring site on Central Yamal, Russia, to test our hypothesis. This area features a high range of ALT (up to 170 cm and locally more), unsaturated moisture conditions over larger areas and shrub tundra as dominating land cover. Tabular ground ice is known to extensively occur on central Yamal [21
]. ALT is measured since 1993 over a CALM grid and several boreholes are in the surrounding area. ALT as well as Mean Annual Air Temperatures (MAAT) have been increasing continuously since the start of the records [21
]. Anomalously high temperatures (and thaw indices) have been observed for 2012 as well as 2016 with unprecedented levels at Mare Sale at least since 1947 [22
]. In situ measurement of subsidence started within the last decade.
Landslides and resulting thermodenudation cirques are typical features on central Yamal [21
] and are also partially represented on the CALM grid. They shape not only terrain but also the distribution of vegetation communities [23
]. Landslides are usually initiated episodically in certain years. All known thermodenudation cirques are regularly surveyed and documented [24
]. The latest year which triggered abundant thermodenudation was 2012 [21
] as it was an exceptionally warm year. Several thermodenudation cirques have been reactivated in 2016 [22
]. A further specific feature of central Yamal is ice-rich permafrost at the base of the active layer. It is in general expected that penetration of thaw into such layers is accompanied by loss of volume due to consolidation and the cause of surface subsidence [25
]. A further feature on central Yamal are deep craters first suggested to be triggered by the extreme conditions in 2012 [26
] and more in 2016. Satellite derived displacement measurements may provide advanced insight into the impact of temperature extremes in combination with in situ measurements.
A range of SAR time series from different missions (including X-, C and L-band) are available for central Yamal, but no acquisitions exist for the year 2012 from for science commonly used missions such as TerraSAR-X, COSMO-Skymed, ALOS (failed in 2011), Sentinel-1 (not yet launched), ENVISAT ASAR (failed in April 2012), Radarsat (ScanSAR only). COSMO-Skymed, Sentinel-1 and ALOS-2 acquisitions are, however, available for 2016 and may allow for assessment of the warmer than average conditions on ground thaw on central Yamal. A constraint is the lack of spatial overlap of COSMO-Skymed and the CALM site what needs to be considered. Furthermore, a regularly year to year connection with C-band is not feasible over this region, as the abundant shrub cover is causing strong decorrelation over one year.
We analyze two years (2013 and 2016) of dense time series of the COSMO-Skymed constellation and all three years so far available from Sentinel-1 (2016, 2017, 2018) over central Yamal. The start of records from COSMO-Skymed coincides with the beginning of soil thaw in both years and for Sentinel-1 in one year what allows both a direct comparison of the X and C-band records and investigation for early season gap filling. We eventually discuss the results obtained in the context of ALT monitoring and implications of temperature extremes as recently recorded on Yamal.
So far, InSAR studies in permafrost environments have focused on areas with little vegetation, with shallow active-layer thickness in the Arctic or the Tibet plateau where limited amount of organic material is present. The active-layer thickness on central Yamal exceeds 100 cm and different types of land cover occur, largely with tundra shrubs. We demonstrate the applicability and added value of InSAR-derived summer displacements for permafrost monitoring in such environments. An important aspect is the scarcity of specific in situ activities tailored to validate the InSAR results. We demonstrate the utility of a simple, but distributed and multi-year measurement setup.
The impact of warmer years on subsidence has been demonstrated before with in situ measurements in the Lena Delta [12
]. Our results confirm increased subsidence also with satellite retrievals. The joint analyses with land cover reveal impact differences with respect to ground properties. Especially the dense time series available from COSMO-Skymed due to its special constellation allows for detailed assessment of thaw progression and impact assessment of extreme years. Results indicate presence and melt of ground ice at the base of the active layer in case of above average air temperature (and subsequent higher degree-days of thawing) conditions. Our results exemplify the impact of higher temperatures and the role of near surface soil properties. Knowledge regarding presence of ground ice is crucial for climate change impact assessment and its mapping one of the present key issues in permafrost research [69
The comparison of the X-band records with C-band based on degree-days of thawing allows for evaluation of displacement retrievals. The agreement between the two sensors and the results from evaluation with in situ data suggest that InSAR-based seasonal subsidence retrievals in shrub dominated tundra are reasonable. The agreement between the wavelengths also suggests that changes of near surface soil water content and subsequent change of signal penetration depth are only of minor importance in such environments. The typical underestimation of displacement by InSAR may result from retrieval issues at the very beginning of thaw since rates agree later on.
We provide a strategy for comparing subsidence observations between years that allows for enhanced understanding of permafrost thaw. However, only L-band preserves long-time coherence over natural terrain, which is required to study long-term evolution of subsidence. Further assessment over similar monitoring sites as on central Yamal, with dense observations of active-layer thickness, are required to develop a scheme for fully exploiting the capabilities of InSAR-derived seasonal subsidence across the entire Arctic.