Tracking land surface deformation in lowland permafrost regions across the Arctic exploiting the first decade of Copernicus Sentinel-1

Synthetic aperture radar interferometry (InSAR)-derived land surface deformation serves as indication for potential permafrost degradation and identification of its drivers. Long-term (multi-annual) as well as seasonal (unfrozen period) deformation are of interest in this context. Primarily data from the first decade of Sentinel-1 (C-band) were investigated but for long-term deformation a combination with PALSAR-2 (Phased Array type L-band Synthetic Aperture Radar-2) observations needed to be considered to account for Sentinel-1 availability limitations and to compile a dataset representing different Arctic lowland permafrost landscapes. Seasonal deformation patterns were found to differ significantly from long-term patterns. Spatial filtering, applied to correct for atmospheric and ionospheric disturbances, led to centring of the density distribution around zero. The resulting positive deformation patterns largely coincided with dry and barren tundra types, which are likely stable areas. A Random Forest (RF) regressor analysis was used to identify driving features for both seasonal and long-term deformation. Considered features included land cover, soil properties, terrain, and disturbances. Both deformation types reflected land cover (vegetation patterns) in addition to disturbances (fires). Mean annual ground temperature was of relatively high importance in case of long-term deformation, for both C- and L-band. The predictive potential was higher for seasonal than for long-term rates, particularly when considering multiple parameters. Parameter importance sequence was more uniform for multi-annual deformation across the analyzed regions. Overall, model fitting scores were only weak to moderate underscoring the need for continuous monitoring with InSAR. Sentinel-1 and PALSAR-2 provided comparable deformation information as well as RF scores which substantiates their combined use.